KR100601729B1 - Spatial inverse filtering device and method considering human cognitive aspect and computer readable recording medium storing computer program controlling the device - Google Patents

Abstract

Disclosed are an apparatus and method for spatial inverse filtering considering a human cognitive aspect and a computer-readable recording medium storing a computer program for controlling the apparatus. The apparatus for filtering first to Nth audio input signals and outputting the first to Nth speakers, respectively, includes first to Nth filters and first to nth filters for filtering first to Nth audio input signals, respectively. An extracting unit for extracting N filtered results by combining the 2N filtered results output from the N th filters, wherein the n (1 ≦ n ≦ N) filter is a space of the n th speaker from the n th audio input signal; A spatial transfer function measuring unit measuring a transfer function, an angle measuring unit measuring an angle of the n th speaker from an n th audio input signal, and a spatial transfer function of the n th speaker, and correcting the n th speaker using the corrected result Characteristic correction unit for correcting the angle of the output to the extraction unit. Therefore, it is possible to achieve the optimal listening conditions recommended by the ITU-T BS 775 by compensating the speaker's direction and compensating the sound quality of the entire band by compensating the non-ideal characteristics of the speaker and the space.

Description

Spatial inverse filtering apparatus and method considering human's perception and computer-readable recording media storing computer program controlling the apparatus}

1 is a diagram for explaining general system equalization.

2 is a block diagram of a spatial inverse filtering apparatus according to the present invention.

3 is a flowchart illustrating a spatial inverse filtering method according to the present invention.

4 is a block diagram according to the present invention of the n-th filter shown in FIG.

FIG. 5 is a diagram for better understanding of the RTF measurement unit illustrated in FIG. 4.

6 is a diagram for aiding in understanding the angle measuring unit when N = 5.

7 (a) and 7 (b) are diagrams for describing a process of obtaining a position of an n-th speaker in the angle measuring unit.

8 is a diagram illustrating an actual appearance of microphones for measuring an angle of an nth speaker.

9 (a) to 9 (c) show the results measured in each channel, that is, in each microphone.

10 is an exemplary diagram for explaining the effect of the apparatus and method for space inverse filtering according to the present invention.

FIG. 11 is a block diagram of an embodiment according to the present invention of the characteristic corrector shown in FIG. 4.

12 is a block diagram of an embodiment according to the present invention of the space transfer function correction unit shown in FIG. 11.

13A to 13D illustrate waveforms input and output to the space transfer function correcting unit illustrated in FIG. 12.

14 exemplarily shows an arrangement of a virtual 5.1 channel speaker.

FIG. 15 is a block diagram of an embodiment of the present invention according to the present invention.

16 (a) and (b) are diagrams to help understanding of the first virtual sound source synthesis unit.

17 is a block diagram of a preferred embodiment of the present invention of the tone correction filter shown in FIG.

18 is a block diagram of an embodiment according to the present invention of a first, second or third virtual sound source synthesis unit.

FIG. 19 is a graph illustrating a distribution according to a stereotactic angle of velocity gains represented in Equations 5 and 6, respectively.

FIG. 20 is a graph illustrating a distribution according to an orientation angle of energy gains represented in Equations 5 and 6, respectively.

FIG. 21 is a block diagram of an embodiment of the first, second, or third virtual sound source synthesizer shown in FIG. 18.

22 is an exemplary view for explaining the effect of the apparatus and method for space inverse filtering according to the present invention.

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the design of a spatial inverse filter used to improve the sound quality of a home theater. In particular, the speaker itself is not ideal and its arrangement in a general listening environment such as a living room. The present invention relates to an apparatus and method for spatial inverse filtering in consideration of a human cognitive aspect capable of compensating for deterioration of sound quality, and a computer-readable recording medium storing a computer program for controlling the apparatus.

The main factors causing the deterioration of sound quality are speaker characteristics, non-uniformity of the distance between the speakers and the listener, characteristics of the listening space, and asymmetry of the speaker's placement angle with respect to the listener. Conventionally, there are many techniques for correcting deterioration of sound quality, but there is a problem in that a sufficient low tone band cannot be corrected due to synchronization problem of image and sound. In addition, since the conventional method does not consider the influence of the speaker arrangement angle, there is a problem that the degradation of the 'sound stage' due to the speaker arrangement angle asymmetry cannot be corrected.

An object of the present invention is to provide a spatial inverse filtering apparatus in consideration of human cognitive aspects that can compensate for the non-ideal characteristics of speakers and space.

Another technical problem to be achieved by the present invention is to provide a spatial inverse filtering method in consideration of human cognitive aspects that can compensate for the non-ideal characteristics of the speaker and the space.

Another object of the present invention is to provide a computer-readable recording medium for storing a computer program for controlling a spatial inverse filtering device in consideration of the human cognitive aspect.

In order to achieve the above object, the spatial inverse filtering apparatus according to the present invention for filtering the first to N-th (where N is a positive integer of 2 or more) and output to the first to N-th speakers, respectively, Extracts N filtered results by combining first through Nth filters that respectively filter the first through Nth audio input signals, and 2N filtered results output by the first through Nth filters. Consists of extracting unit to

The n th (1 ≦ n ≦ N) filter measures a space transfer function measurement unit for measuring a space transfer function of the n th speaker from an n th audio input signal, and measures an angle of the n th speaker from the n th audio input signal. Preferably, the angle measuring unit and the spatial transfer function of the n-th speaker are corrected, and the characteristic correction unit corrects the angle of the n-th speaker using the corrected result and outputs the extracted part to the extraction unit.

In order to achieve the above object, the first to the N-th (where N is a positive integer of 2 or more). The spatial inverse filtering method according to the present invention for filtering the audio input signals and output to the first to N-th speakers, respectively Filtering each of the first to Nth audio input signals and extracting N filtered results by combining 2N filtered results,

Filtering the n th (1 ≦ n ≦ N) audio input signal comprises measuring a spatial transfer function of the n th speaker and an angle of the n th speaker from an n th audio input signal and the spatial transfer of the n th speaker Compensating a function, and correcting the angle of the n-th speaker using the corrected result.

In order to achieve the above object, the first to the Nth (where N is a positive integer of 2 or more) to control the spatial inverse filtering device for filtering the audio input signals and output to the first to Nth speakers, respectively In a computer-readable recording medium according to the present invention for storing a computer program, the computer program comprises filtering the first through the N-th audio input signals and combining the 2N filtered results into N filtered results. Take the steps to extract the results,

Filtering the n th (1 ≦ n ≦ N) audio input signal comprises measuring a spatial transfer function of the n th speaker and an angle of the n th speaker from an n th audio input signal and the spatial transfer of the n th speaker Compensating a function, and correcting the angle of the n-th speaker using the corrected result.

Prior to describing the present invention, a conventional general spatial equalization technique is described as follows. For spatial equalization techniques, titled "Discrete-Time Signal Processing", "Oppenheim, A. V." And a book by Schafer, AW, published in 1989 by Prentice-Hall, Englewood Cliffs, NJ, and entitled "Adaptive Filter Theory", by Haykin, S., Prentice-Hall, Englewood 3rd edition, published in 1996 by Cliffs, NJ.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram for explaining general system equalization, a room inverse filter [H (z)] 10 and a room transfer function (RTF) [G (z). ] (12).

In the general case, the input signal x [n] is distorted due to the space transfer function G (z), and a space inverse filter H (z) 10 is provided to compensate for this distortion. If the space inverse filter [H (z)] (10) is theoretically perfect, then the product of the space transfer function [G (z)] and the space inverse filter [H (z)] becomes a unit impulse function as Can be.

However, since the space transfer function [G (z)] actually includes various phase characteristics and the area that the speaker can reproduce is generally limited, instead of '1' on the right side of Equation 1 as shown in Equation 2 below. A predetermined objective function [T (z)] is adopted.

Here, E (z) is an error function representing an error expressed as in Equation 3 below.

In general, in order to obtain H (z), T (z) is set to have the same phase characteristic as G (z) and a magnitude characteristic of '1' in the frequency region where equalization is desired, and E (z) is a predetermined reference. Obtained by the method of minimization by criterion. For this purpose, the implementation of H (z), which is mainly used for this, is a band equalizer by an infinite impulse response (IIR) filter and a magnitude characteristic equalizer by a linear phase finite impulse response (FIR) filter. Etc. In addition, there is an inverse phase FIR filter that compensates for the phase characteristic of G (z) by an adaptive filter implementation technique, but this is rarely used because of the excessive computational amount and the difficulty of implementation. Here, three representative conventional general techniques for obtaining H (z) are described.

First, the technique using the band equalizer by the IIR filter divides the spatial transfer function into several bands, finds the energy contained in each band, and then obtains the gain for each band to equalize the energies. In this case, after the target frequency magnitude response curve is set, inverse filtering is performed in a manner that compensates for the difference in average energy for each octave band obtained from the frequency magnitude response curve for the set curve and the spatial impulse response. Since this technique can be configured using a simple IIR filter such as a Biquad filter, its implementation is simple and is most widely used to obtain H (z). However, this technique has a limitation in that it is not possible to compensate for a sudden change in loudness occurring in the same band, that is, it cannot compensate for the minute characteristics appearing in the band, and large phase distortion occurs in an area where two bands overlap. Has a problem.

Next, the technique using the magnitude characteristic equalizer by the linear phase FIR filter obtains the magnitude characteristic of the spatial transfer function, and has the magnitude response and the linear phase response divided by the objective function as This is how to get the FIR filter.

This conventional technique has the advantage of compensating for relatively fine frequency characteristics without additional phase distortion. However, since the implementation is complicated due to a large amount of calculation, there are not many examples of actual products. The frequency resolution that can be obtained with this technique is proportional to the length of the overall filter, but if you increase it too much, the delay of the sound caused by this filtering will be large, and according to the purpose, appropriate adjustment is necessary. In order to utilize it, it is important to utilize the lip sync as long as it does not hurt. In addition, this technique can provide finer compensation than an IIR filter-based band equalizer, but it requires a large number of taps to finely compensate the characteristics of the low range, which is not suitable for fine compensation of the low range. . That is, this conventional technique has a problem of requiring a large amount of computation in order to obtain sufficient resolution up to the low frequency band.

Finally, the technique using the spatial mode equalizer by the IIR filter reduces the resonance mode of the space caused by the most prominent problem in the space transfer function. First, the space transfer function is obtained, and then the resonance mode of the space is measured and a notch filter corresponding to the space is measured to reduce the resonance mode. The technique of using the band equalizer by the IIR filter and the size characteristic equalizer by the linear phase FIR filter described above solve the problems of not only the spatial characteristics but also the speaker's misplacement and the speaker's incomplete frequency characteristics. On the other hand, the technique using the spatial mode equalizer by the IIR filter only compensates for the defects in sound transmission which are inevitable in the closed space.

As a result, the above-mentioned conventional techniques all have a problem in that the 'sound stage' degradation, which is most important for stereoscopic listening caused by an undesired speaker placement angle, cannot be corrected.

Hereinafter, with reference to the accompanying drawings, the configuration and operation of the spatial inverse filtering apparatus in consideration of the human cognitive aspect and the spatial inverse filtering method performed in the apparatus according to the present invention will be described as follows.

FIG. 2 is a block diagram of a spatial inverse filtering apparatus according to the present invention, and includes first to N-th filters 22, 24,..., And 26 and an extractor 30. Here, N is a positive integer of 2 or more.

3 is a flowchart illustrating a spatial inverse filtering method according to the present invention, comprising: filtering each audio input signal (steps 40 and 42) and extracting only some of the filtered results (44) Step).

The apparatus and method for spatial inverse filtering according to the present invention filters the first to N th audio input signals and outputs the first to N th speakers (not shown), respectively.

First, in the first to the N-th filter shown in Figure 2 (22, 24, ..., and 26) an input terminal IN _1, IN _2, ..., and first to N respectively input through the IN _N The audio input signals are respectively filtered and the filtered result is output to the extractor 30 (40th and 42nd steps). At this time, the operation of each filter is as follows.

FIG. 4 is a block diagram according to the present invention of the n-th filter shown in FIG. 2 and includes a space transfer function (RTF) measuring unit 50, an angle measuring unit 52, and a characteristic correcting unit 54. Here, 1≤n≤N.

The spatial transfer function measuring unit 50 of the n th filter measures the spatial transfer function of the n th speaker from the n th audio input signal input through the input terminal IN _n , and measures the spatial transfer function of the n th speaker. Output to the government 54 (step 40). Here, the spatial transfer function of the n-th speaker refers to the frequency characteristic of the path through which sound is transmitted from the n-th speaker to the position of the listener. Here, the frequency characteristic includes information about the loudness and the time at which the sound is delivered. For the space transfer function, titled "Hundamentals of Acoustics". It is published in Kuttruff's fourth edition, published in 2000 by Wiley & Sons. Signals used to measure the space transfer function include maximum length sequence (MLS), impulse, sine sweep, and so on. You may use it.

FIG. 5 is a view for better understanding of the RTF measurement unit 50 shown in FIG. 4. The n-th speaker 60 and the microphone 62 provided in the room 64 are illustrated. Indicates. In FIG. 5, although only one microphone 62 is shown, multiple microphones may be provided.

Referring to FIG. 5, sound leaked through the n-th speaker 60 may be directly received by the microphone 62 or reflected by the room 64. In this case, the space transfer function measuring unit 50 performs fast Fourier transform of a room impulse response measured using a sound received by the microphone 62 on the assumption that several microphones are arranged in close proximity to each other. (FFT: Fast Fourier Transform) to measure the space transfer function.

In operation 40, the angle measuring unit 52 measures the angle of the n th speaker from the n th audio input signal input through the input terminal IN _n, and outputs the measured angle to the characteristic correction unit 54. . Here, the angle of the n-th speaker means the arrangement angle of the n-th speaker from the position of the listener, that is, the angle between the position of the speaker and the listener arranged in a specific space.

FIG. 6 is a diagram for better understanding of the angle measuring unit 52 when N = 5, and includes five speakers 70, 72, 74, 76, and 78 and three microphones 80, 82, and 84. It consists of.

Referring to FIG. 6, the angle measuring unit 52 may measure the angle of the n-th speaker 70, 72, 74, 76, or 78 using three microphones 80, 82, and 84.

Hereinafter, the process of obtaining the position of the n-th speaker to measure the angle of the n-th speaker by the angle measuring unit 52 will be described as follows with reference to the accompanying drawings.

7 (a) and 7 (b) are diagrams for describing a process of obtaining a position of an n th speaker in the angle measuring unit 52. In FIG. 7B, c represents a sound velocity and t represents time.

8 is a diagram illustrating an actual appearance of microphones for measuring an angle of an nth speaker.

9 (a) to 9 (c) show the results measured in each channel, that is, in each microphone, wherein the vertical axis of each graph represents an amplitude and the horizontal axis represents the number of samples, respectively.

In order to measure the position of the speaker consisting of the distance and angle of the n-th speaker centered on the listener, the spatial impulse response is simultaneously measured on three different channels. For each channel, time difference t _o , t _x and t _y from the measurement start time point (indicated by '0' in FIG. 7 (b)) to the time when a direct sound appears on the impulse response is shown in FIG. 7 (b). Can be obtained as shown in From the information on these time differences t _o , t _x and t _y , the distances R ₀ , R _X and R _Y from the measuring point to the speaker can be obtained. The n-th speaker exists on an arc whose radius is the distance R ₀ , R _X and R _Y corresponding to the time differences t _o , t _x and t _y calculated around the measuring point. At this time, the intersection point 86 of the three circular arcs as shown in Fig. 7 (a) satisfying all the time delays obtained from the three channels ch1, ch2 and ch3 is uniquely defined in space and the point ( 86) is the position of the n-th speaker. To this end, the microphones may be arranged as shown in FIG. 8 to directly detect sound as shown in FIGS. 9A to 9C.

After the 40th step, the characteristic corrector 54 corrects the spatial transfer function of the n-th speaker input from the RTF measuring unit 50, corrects the angle of the n-th speaker using the corrected result, and corrects the result. Is output to the extraction unit 30 through the output terminal OUT _n (step 42).

After the 42 th step, the extractor 30 extracts N filtered results by combining 2N filtered results output from the first to Nth filters 22, 24,. The filtered N filtered results are output through the output terminals OUT ₁ , OUT ₂ ,..., And OUT _N , respectively (step 44). Here, each filter 22, 24,... Or 26 filters the audio input signal by dividing the audio input signal into right and left components, so that the components output from each filter 22, 24,. Dog.

FIG. 10 is an exemplary diagram for explaining the effect of the apparatus and method for spatial inverse filtering according to the present invention, and the virtual positions and six speakers 90, 92, 94, 96, 98, and 100 existing in actual positions. Six speakers 110, 112, 114, 116, 118 and 120 present in the FIG.

When N is 6, referring to FIG. 10, the signal output from the extractor 30 through the output terminal OUT ₁ should be located at the front right (FR: Front Right) centered on the listener 102. ) And the signal output from the extraction unit 30 through the output terminal OUT ₂ is input to the speaker 92 to be located in the center (C: Center) in front of the listener 102, the output terminal The signal output from the extractor 30 through OUT ₃ is input to a subwoofer (SW) 94 which should be located on the front left side around the celadon 102 and the extractor through the output terminal OUT ₄ . The signal output from the 30 is input to the speaker 96 which should be located on the front left (FL: Front Left) centered on the celadon 102, and the signal output from the extraction unit 30 through the output terminal OUT ₅ is a rear left around the listener (102) is input to the speaker (98) to be located at (RL rear left), through the output terminal OUT ₆ Signal output from chulbu 30 rear right around the listener (102) is input to the speaker 100 to be located in the (RR Rear Right). In this case, the first to sixth audio input signals are filtered and extracted by the extraction unit 30 and input to the speakers 90 to 100, respectively, by the spatial inverse filtering apparatus and method according to the present invention. ) Can listen to the sound as if the speaker is placed at the virtual positions 110-120 to generate sound.

FIG. 11 is a block diagram of an embodiment 54A according to the present invention of the characteristic corrector 54 shown in FIG. 4, which is composed of a space transfer function corrector 130 and an angle corrector 132.

The space transfer function correction unit 130 corrects the space transfer function of the n-th speaker input through the input terminal IN1, and outputs the corrected result to the angle correction unit 132. In most cases, serious problems of the space transfer function appear in the low frequency range. When the space transfer function is corrected by the space transfer function correcting unit 130, the characteristics of the space and the speaker itself, for example, frequency characteristics and ideal It is possible to compensate for the deterioration of sound caused by non-speaker arrangement.

The angle corrector 132 corrects the angle of the n-th speaker input through the input terminal IN2 from the result corrected by the space transfer function corrector 130, and outputs the corrected result through the output terminal OUT1.

FIG. 12 is a block diagram of an embodiment 130A according to the present invention of the spatial transfer function correcting unit 130 shown in FIG. 11, wherein the first octave band smoothing unit 140 and the full band filter ( 142 and a resonance mode cancellation filter 144.

The first octave band smoothing unit 140 shown in FIG. 12 performs a predetermined ratio, for example, 1/3 band octave smoothing on the spatial transfer function of the n-th speaker input from the RTF measurement unit 50 through the input terminal IN3. Reduce drastic changes in frequency characteristics. This is referred to as "Generalized Fractional-Octave Smoothing of Audio and Acoustic Responses." Hatziantoniou 'and' J.N. Mourjopoulos' and published in a 2000 article in the Journal of Audio Engineering Society., Vol 48, No.4.

The full band filter 142 is a linear phase FIR filter having the inverse of the output of the first octave band smoothing unit 140 as its magnitude characteristic, and filters the output of the first octave band smoothing unit 140. To this end, according to the present invention, the full band filter 142 may be implemented as a low frequency FIR filter 150 and a high frequency FIR filter 152.

Here, the low frequency FIR filter 150 inversely filters the low frequency band of the output of the first octave band smoothing unit 140 by a finite impulse response (FIR) and outputs the result to the resonance mode cancellation filter 144. do. The high frequency FIR filter 152 filters the high frequency band of the output of the first octave band smoothing unit 140 by the finite impulse response inverse, and outputs the result to the resonance mode cancellation filter 144. As described above, the full band filter 142 is divided into low frequency bands and high frequency bands, and thus higher resolution can be given to the low frequency area.

The resonance mode cancellation filter 144 corrects the resonance mode of the space transfer function from the output of the full band filter 142, and outputs the corrected result to the angle correction unit 132 through the output terminal OUT2. To this end, the resonance mode cancellation filter 144 is implemented as a plurality of biquad notch filters that eliminate the spatial resonance mode appearing at the output of the full-band filter 142, so that the resonance mode occurring in the listening space Can reduce them. This is referred to as "Digital Filters" under the heading 'A. It is published in the second edition, written by Antonius' and published in 2000 by McGraw-Hill.

At this time, the upper limit frequency of the resonance mode cancellation filter 144 may be determined by the Schroeder frequency determined by the size of the space. In this regard, entitled "Fundamentals of Acoustics", the 'H. It is published in Kuttruff's fourth edition, published in 2000 by Wiley & Sons. This Schröder frequency is used to estimate the distance between the measured microphone and speakers. This inaccurate but practically practical approximation of Schröder frequency can be used.

At this time, the frequency-specific volume of the full-band filter 142 and the resonance mode cancellation filter 144 shown in FIG. 12 in order to prevent loss of excessive dynamic range and nonlinear distortion when reducing the resonance modes. The variation may be limited to a certain level. Here, the predetermined level may be -10 decibels to 3 decibels.

13A to 13D illustrate waveforms input / output to the space transfer function correcting unit 130A shown in FIG. 12. In FIG. 13 (c), the thin solid line represents the signal output from the space transfer function correcting unit 130A, the dashed-dotted line represents the signal input to the space transfer function correcting unit 130A, and the dotted line represents the biquad notch filter 1 ( shows the frequency response (FR) of biquad # 1), the dashed line shows the frequency response (FR) of biquad notch filter 2 (biquad # 2), and the thick solid line shows the biquad notch filter 3 (biquad # 3). Denotes the frequency response (FR) of each.

13 (c) shows waveform diagrams when the resonance mode cancellation filter 144 is composed of three biquad notch filters. At this time, each of the biquad notch filters performs suppression for the resonance mode one by one. The dashed-dotted line shown in FIG. 13 (c) shows frequency characteristics of a 0 to 150 kHz band of a signal output when a signal having a flat magnitude level is applied to all frequency bands, and a solid line shows a resonance mode cancellation filter 144. It shows the result of changing the above-described frequency characteristics when passing through.

As a result, the present invention imparts distortion opposite to that caused in the actual environment, so that the distortions are actually canceled out.

When the spatial transfer function is as shown in Fig. 13A, the full band filter 142 equalizes the full frequency band components as shown in Fig. 13B. In this case, the full band filter 142 does not have sufficient resolution to remove the resonance mode of the space in the low frequency band. Therefore, the resonance mode cancellation filter 144 is provided. The resonance mode cancellation filter 144 suppresses the resonance of the low frequency band shown in FIG. 13A in the direction of the arrow as shown in FIG. 13C.

As a result, in the process of spatial correction, the main band filter 142 plays a main role, and the resonance mode cancellation filter 144 plays a secondary role. The full band filter 142 also corrects even the lowest frequency. However, due to various circumstances, it is difficult to precisely correct in a low frequency band, and thus a resonance mode cancellation filter 144 is provided to compensate for this.

 In this case, a waveform as shown in FIG. 13D may be output from the space transfer function corrector 130 or 130A.

FIG. 14 is a diagram illustrating an exemplary arrangement of a virtual 5.1 channel speaker, which is represented by the speakers 160, 162, 164, 166, and 168 that exist at actual positions.

When the space transfer function is corrected by the space transfer function correcting unit 130 or 130A as shown in FIG. 11 or 12, it is generated from the speakers 160, 162, 164, 166, and 168 present in the actual position. The sound being heard may be heard to the listener 167 as if it were generated from speakers 161, 163, 165, 166 and 169 present at virtual positions in the direction of the arrow shown in FIG. 14.

In general, in order to correct the low frequency band of the space transfer function, it is necessary to design a filter having a very high resolution, which causes a high calculation amount and a delay time due to signal processing. However, when the space transfer function correcting unit 130 is implemented as shown in FIG. 12, the low frequency band of the space transfer function can be effectively processed without causing such problems.

Meanwhile, according to the present invention, the angle corrector 132 illustrated in FIG. 11 may correct the angle of the nth speaker to an angle recommended by the ITU-T BS775. To this end, the angle correction unit 132 uses a panning technique and uses the virtual loudspeaker position (VLP) to determine the angle of the n th speaker input through the input terminal IN2 from the result corrected by the space transfer function correction unit 130. It can also be corrected by using the Virtual Loudspeaker Positioning method. This is written by 'Sin-Lyul Lee', 'Ki-Young Han', 'Seung-Rae Lee' and 'Koeng-Mo Sung' under the heading "Enhanced Constant Power Panning Law for Surround Sound Systems", Proceedings of ICA 2004, VI-3155 to V1-3156. The VLP method is a kind of constant power panning (CPP) technique. It is known that the distortion of sound quality is small and the most accurate virtual sound source positioning is possible. This CPP technique is known as' H. Wireless World, published in 1968 by D. Harwood, entitled "Stereophonic image sharpness." vol. 74, pages 207-211. Or Master's thesis, published in 1998 by West J. R of the School of Music Engineering at the University of Miami, entitled "Five-channel panning laws: an analytic and experimental comparison." ) Is disclosed. In addition to the VLP method, the angle correction unit 132 may use techniques such as sine-cosine (CPP) CPP or Head Related Transfer Function (HRTF) positioning.

Hereinafter, the configuration and operation of a preferred embodiment of the angle correction unit 132 using the VLP method according to the present invention will be described with reference to the accompanying drawings.

FIG. 15 is a block diagram of an embodiment 132A according to the present invention of the angle corrector 132 shown in FIG. 11, and includes a first virtual sound source synthesizer 170 and a sound color correction filter (SCC). 172.

Referring to FIG. 15, the first virtual sound source synthesizing unit 170 of the angle corrector 132A may consider gains determined by using the angle of the n th speaker input through the input terminal IN4 in consideration of the human psychological factor. The high and low frequency components separated from the corrected result input from the space transfer function corrector 130 through the input terminal IN5 are calculated, and the calculated result is output to the tone correction filter 172.

16 (a) and (b) are diagrams to help understanding of the first virtual sound source synthesizer 170, and FIG. 16 (a) shows speakers 174, 176, 178, centered on the listener 173. 181 and 182, and FIG. 16 (b) shows the speakers 183, 184, 185, 186 and 187 around the listener 189. Here, the arrow in Fig. 16 (a) indicates that the psychological position of the speaker is moved by positioning.

As a result, by the first virtual sound source synthesizing unit 170 shown in FIG. 15, the position of the virtual sound source is located in a straight line between two speakers used for positioning as shown in FIG. Therefore, time delay is added again to adjust the distance between each speaker 183, 184, 185, 186, 187 and the listener 189 in the direction of the arrow as shown in FIG. 16 (b).

On the other hand, when the position of the n-th speaker is changed by the first virtual sound source synthesizing unit 170, since the physical arrangement between the n-th speaker and the listener's ear is different, the HRTF generated by the listener's head and torso is Will be different. Since the VLP method is not a stereotactic technique using the HRTF, it is necessary to correct the tone caused by the HRTF difference. In this way, the tone correction filter 172 calculates the difference between the HRTF at the actual position of the n-th speaker and the HRTF at the virtual position, and serves to correct the tone as much as the difference. To this end, the tone correction filter 172 is input through the input terminal IN4 by the difference between the HRTFs at the virtual position and the virtual position of the nth speaker by using the result calculated by the first virtual sound source synthesis unit 170. The angle of the speaker is corrected, and the corrected result is output to the extraction unit 30 through the output terminal OUT3.

FIG. 17 is a block diagram of a preferred embodiment 172A according to the present invention of the tone correction filter 172 shown in FIG. 15, wherein the first, second and third HRTF calculation units 190, 192, and 194 are shown. To the second and third virtual sound source synthesis units 196 and 198, the adder 200, the second and third octave band smoothing units 202 and 204, the component difference calculator 206 and the size inverter 208. It is composed.

The first HRTF calculator 190 calculates an HRTF at the virtual position of the n-th speaker and outputs the calculated result to the second octave band smoothing unit 202. The second HRTF calculator 192 calculates an HRTF at an actual left position of the n-th speaker by using the measured position of the n-th speaker input through the input terminal IN6, and calculates the HRTF at the second left virtual sound source synthesizer. Output to (196). The third HRTF calculating unit 194 calculates an HRTF at an actual right position of the n-th speaker by using the measured position of the n-th speaker input through the input terminal IN6, and calculates the HRTF at the actual right position of the n-th speaker. Output to (198).

The second virtual sound source synthesis unit 196 may take into account gains determined by using the angle of the n-th speaker input from the angle measuring unit 52 through the input terminal IN6 and considering the human psychological factor, and the second HRTF calculating unit ( The high and low frequency components separated from the result calculated in 192 are calculated, and the calculated result is output to the adder 200.

The third virtual sound source synthesizing unit 198 considers the human psychology factor and uses the gains determined using the angle of the n-th speaker input from the angle measuring unit 52 through the input terminal IN6 and the third HRTF calculating unit ( The high and low frequency components separated from the result calculated in 194 are calculated, and the calculated result is output to the adder 200.

The adder 200 adds the results calculated by the second and third virtual sound source synthesizers 196 and 198, and outputs the added result to the third octave band smoothing unit 204.

After all, since one virtual speaker is implemented using two real speakers, for this purpose, the second and third of the HRTFs corresponding to the real speaker pairs and calculated by the second and third HRTF calculators 192 and 194 are used. The results calculated by the virtual sound source synthesizers 196 and 198 are respectively synthesized by the adder 200 to obtain a result of one HRTF.

The second octave band smoothing unit 202 outputs the result calculated by the first HRTF calculating unit 190, for example, 1/3 octave band smoothing, and outputs the octave band smoothing result to the component difference calculating unit 206. do. In addition, the third octave band smoothing unit 204, for example, 1/3 octave band smoothing of the result added by the adding unit 200, and outputs the octave band smoothing result to the component difference calculating unit 206. The component difference calculator 206 calculates a difference between the output components of the second and third octave band smoothing units 202 and 204, and outputs the calculated difference to the magnitude inversion unit 208.

The magnitude inversion unit 208 has a characteristic determined by the inverse of the difference calculated by the component difference calculator 206, and filters the result calculated by the first virtual sound source synthesis unit 170 through the input terminal IN7. The filtered result is output through the output terminal OUT4 as a result of correcting the angle of the n-th speaker. Here, the magnitude inversion unit 208 may be implemented as an FIR filter.

Hereinafter, with reference to the accompanying drawings, the configuration and operation of the preferred embodiment according to the present invention of each of the first, second and third virtual sound source synthesizers 170, 196 and 198 shown in Figs. Explain together.

FIG. 18 is a block diagram of an embodiment of the first, second or third virtual sound source synthesizing unit 170, 196 or 198 according to the present invention, wherein the component separating unit 220, the gain generating unit 222, and the multiplication unit ( 224 and the adder 226.

When the embodiment illustrated in FIG. 18 corresponds to the first virtual sound source synthesizing unit 170, the component separating unit 220 may input a result corrected by the spatial transfer function correcting unit 130 through the input terminal IN8 to provide a predetermined result. The low frequency component and the high frequency component are separated based on the reference frequency, and the separated high frequency component and the low frequency component are output to the multiplier 224.

However, when the embodiment illustrated in FIG. 18 corresponds to the second or third virtual sound source synthesizer 196 or 198, the component separator 220 may calculate the second or third HRTF calculator 192 or 194. The result is input through the input terminal IN8 to separate the low frequency component and the high frequency component based on the predetermined reference frequency, and output the separated high frequency component and the low frequency component to the multiplier 224. The reference frequency is previously determined as a frequency dividing the directional psychological factor in the frequency band, and may be 700 Hz, and the reference frequency component may be determined to belong to the high frequency component, or may be determined to belong to the low frequency component. Here, for the detailed operation of the component separation unit 220, 92nd Convention of the Audio Engineering Society, J. Audio, published in May 1992 under the title "General metatheory of auditory localization" by Michael A. Gerzon. Eng. Soc., Preprint 3306.

The gain generator 222 considers the human psychological factor and uses the determined energy vector direction and velocity vector direction and the angle of the n-th speaker input through the input terminal IN9, The speed gain is generated, and the generated energy gain and the speed gain are output to the multiplier 224. The study of human directional psychological factors was organized by Gerzon. Gerzon introduced the concept of velocity vectors, energy vectors, and phasiness to optimize panning rules. See, for example, the 92nd Convention of the Audio Engineering Society, J. Audio Eng., Published in May 1992, entitled "Panpot laws for multispeaker stereo" by Michael A. Gerzon. Soc., Preprint 3309, or 93rd Convention of the Audio Engineering Society, J. Audio Eng., Published in October 1992, entitled "Psychoacoustic decoders for multispeaker stereo and surround sound" by Michael A. Gerzon. Soc., Preprint 3406. Among the above-mentioned concepts, it was confirmed through subjective evaluation that the velocity vector and the energy vector play the most important role in the phase alignment in the band of less than 700 Hz and the band of 700 Hz or more, respectively. According to the exemplary embodiment of the present invention using this, the gain generator 222 may generate the energy gain g _nE and the speed gain g _nV as shown in Equation 5 below.

Where g _nE represents the energy gain, g _nV represents the speed gain, θ ₁ represents the angle between the position of the measurement microphone and the n-th speaker positioned at the position where the user is expected to be located, and θ ₂ represents the user It represents an angle between the speaker adjacent the measuring microphone and the n speakers placed at the expected location to be present, θ _V ^target and θ _E ^target is θ ₁ and θ ₂ indicated by the direction and measurement results of the sound images the user wants Denotes a target value of the velocity vector direction and a target value of the energy vector direction which are preset using

Alternatively, the gain generator 222 may generate an energy gain and a speed gain as shown in Equation 6 below.

FIG. 19 is a graph illustrating a distribution according to a panning angle of speed gains represented by Equations 5 and 6, respectively, in which the horizontal axis represents the stereotactic angle and the vertical axis represents the gain. Here, g ₁ represents the speed gain expressed in Equation 5, and g ₂ represents the speed gain expressed in Equation 6, respectively.

FIG. 20 is a graph showing a distribution according to an orthogonal angle of energy gains g1 and g2 represented in Equations 5 and 6, respectively, in which the horizontal axis represents the stereotactic angle and the vertical axis represents the gain, respectively. Here, g ₁ represents the energy gain expressed in Equation 5, and g ₂ represents the energy gain expressed in Equation 6, respectively.

As a result, the gains represented in Equations 5 and 6 are distributed according to the stereotactic angle as shown in FIGS. 19 and 20.

The multiplier 224 multiplies the energy gain and the speed gain by the high frequency component and the low frequency component input from the component separator 220, and outputs the multiplied results to the adder 226.

As described above, the reason why the different gain values are applied to the low frequency component and the high frequency component is that human orientation recognition factors show different aspects in the low frequency band and the high frequency band based on about 700 Hz.

The adder 226 adds the results multiplied by the multiplier 224, and outputs the added result through the output terminal OUT5. Here, the adder 226 performs a function of reconstructing and outputting the processed signal divided into a low frequency band and a high frequency band into signals of all bands.

FIG. 21 is a block diagram of an embodiment of the first, second, or third virtual sound source synthesizing unit 170, 196, or 198 shown in FIG. 18, and includes a gain generator 240, a component separator 242, and a multiplier. 244 and an adder 246.

The gain generator 240, the component separator 242, the multiplier 244, and the adder 246 illustrated in FIG. 21 may include the gain generator 222, the component separator 220, It corresponds to the multiplier 224 and the adder 226, respectively, and performs the same function.

The component separator 242 illustrated in FIG. 21 includes first and second low pass filters (LPFs) 260 and 264 and first and second high pass filters (HPFs). 262 and 266. The multiplier 244 is composed of first, second, third and fourth multipliers 270, 272, 274 and 276. The adder 246 is composed of adders 280 and 282.

The first low pass filter 260 of the component separating unit 242 shown in FIG. 21 separates a component less than a predetermined reference frequency as a low frequency component from a signal input through the input terminal IN11, and separates the separated low frequency. The component is output to the first multiplier 270. The first high pass filter 262 separates a component having a predetermined reference frequency or more from the signal input through the input terminal IN11 into a high frequency component, and outputs the separated high frequency component to the second multiplier 272.

In this case, the gain generator 240 generates the energy gain g _1E and the speed gain g _1V using the energy and velocity vector directions determined in consideration of the human direction psychological factor, as shown in Equation 5 above. The generated speed gain g _1V is output to the first multiplier 270, and the generated energy gain g _1E is output to the second multiplier 272. To this end, the gain generator 240 may input θ ₁ , θ ₂ , θ _V ^target and θ _E ^target described in Equation 5 through the input terminal IN10.

The first multiplier 270 of the multiplier 244 multiplies the low frequency component input from the first LPF 260 of the component separation unit 242 by the speed gain g _1V input from the gain generator 240. The multiplied result is output to the adder 280. The second multiplier 272 multiplies the high frequency component input from the first HPF 262 of the component separator 242 by the energy gain g _1E input from the gain generator 240 and multiplies the multiplied result. Output to the adder 280.

The adder 280 of the adder 246 adds the results multiplied by the first and second multipliers 270 and 272, and outputs the added result through the output terminal OUT6.

Similarly, the second low pass filter 264 of the component separator 242 separates a component lower than a predetermined reference frequency as a low frequency component from a signal input through the input terminal IN11, and separates the separated low frequency component. Output to third multiplier 274. The second high pass filter 266 separates a component having a predetermined reference frequency or more from the signal input through the input terminal IN11 into a high frequency component, and outputs the separated high frequency component to the fourth multiplier 276.

In this case, the gain generator 240 generates the energy gain g _2E and the speed gain g _2V using the energy and velocity vector directions determined in consideration of the human direction psychological factor, as shown in Equation 6 above. The generated speed gain g _2V is output to the third multiplier 274, and the generated energy gain g _2E is output to the fourth multiplier 276. To this end, the gain generator 240 may input θ ₁ , θ ₂ , θ _V ^target and θ _E ^target described in Equation 6 through the input terminal IN10.

The third multiplier 274 of the multiplier 244 multiplies the low frequency component input from the second LPF 264 of the component separator 242 by the speed gain g _2V input from the gain generator 240. The multiplied result is output to the adder 246. The fourth multiplier 276 multiplies the high frequency component input from the second HPF 266 of the component separator 242 by the energy gain g _2E input from the gain generator 240 and multiplies the multiplied result. It outputs to the adder 246.

The adder 282 of the adder 246 adds the results multiplied by the third and fourth multipliers 274 and 276, and outputs the added result through the output terminal OUT7.

22 is an exemplary view for explaining the effect of the apparatus and method for space inverse filtering according to the present invention. Here, RL, FL, C, FR and RR represent actual positions of the speakers, and RL_p, FL_p, C_p, FR_p and RR_p represent virtual positions of the speakers recognized by the listener 300. Referring to FIG. 22, when using the angle corrector 132 or 132A illustrated in FIG. 11 or 15, sound generated from speakers located at actual positions in the RL, FL, C, FR, and RR directions may be heard. 300 is perceived as occurring in speakers located in a virtual position moved in the direction of the arrow.

Meanwhile, the above-described embodiments of the present invention may be embodied as computer readable codes / instructions / programs, and the codes / instructions may be implemented using a medium, for example, a computer readable recording medium. / Can be implemented in a general-purpose digital computer for operating the program. The computer-readable recording medium may be a magnetic storage medium (eg, ROM, floppy disk, hard disk, magnetic tape, etc.), optical reading medium (eg, CD-ROM, DVD, etc.) and carrier wave (eg Storage media, such as through the Internet). In addition, embodiments of the present invention may be implemented as a medium (s) containing computer readable code, such that a plurality of computer systems connected through a network may be distributed and processed. Functional programs, codes and code segments for realizing the present invention can be easily inferred by programmers in the art to which the present invention belongs.

As described above, the apparatus and method for spatial inverse filtering in consideration of the human cognitive aspect of the present invention and the computer-readable recording medium storing the computer program for controlling the apparatus are different from the conventional methods. It has a low delay time that can ensure the synchronization of the sound quality, it is possible to precisely compensate for the degradation of the sound quality occurring in the low frequency band, and has the effect of correcting the degradation of the sound quality caused by the problem of the speaker placement angle. That is, the present invention compensates the non-ideal characteristics of the speaker and the space, for example, the frequency transfer characteristics, and compensates the sound quality of the entire band while comprehensively performing the direction correction of the speaker to thereby optimize the optimal listening conditions recommended by the ITU-T BS 775. Has the effect to achieve.

Claims (20)

A spatial inverse filtering device for filtering audio input signals and outputting the first through Nth filters, wherein N is a positive integer greater than or equal to 2; First to Nth filters for filtering the first to Nth audio input signals, respectively; And An extraction unit for extracting N filtered results by combining 2N filtered results output from the first to Nth filters, The nth (1≤n≤N) filter A space transfer function measuring unit configured to measure a space transfer function of the nth speaker from an nth audio input signal; An angle measuring unit measuring an angle of the nth speaker from the nth audio input signal; And A space considering a cognitive aspect of a human being, characterized by correcting a spatial transfer function of the n-th speaker and correcting an angle of the n-th speaker using the corrected result and outputting the extracted feature to the extraction unit. Reverse filtering device. The method of claim 1, wherein the characteristic correction unit A space transfer function correcting unit configured to correct a space transfer function of the n-th speaker; And And an angle corrector configured to correct an angle of the n-th speaker from the corrected result. The method of claim 2, wherein the space transfer function corrector A first octave band smoothing unit for smoothing the spatial transfer function of the nth speaker by a predetermined ratio band octave band; A full band filter having an inverse of the output as a magnitude characteristic in the first octave band smoothing unit; And And a spatial inverse filtering device considering a cognitive aspect of a human being having a resonance mode cancellation filter for correcting a resonance mode of the spatial transfer function from an output of the full band filter. 4. The filter of claim 3, wherein the full band filter is A low frequency FIR filter for inversely filtering a low frequency band of an output to the first octave band smoothing unit; And And a high frequency FIR filter for performing a finite impulse response inverse filtering on the high frequency band of the output in the first octave band smoothing unit. The filter of claim 3, wherein the resonance mode cancellation filter And a biquad notch filter configured to biquad-notch filter the output of the full-band filter. The apparatus of claim 3, wherein the upper limit frequency of the resonance mode cancellation filter is determined by a Schroeder frequency. The apparatus of claim 3, wherein the predetermined ratio is 1/3. The apparatus of claim 3, wherein the frequency variation of the full band filter and the resonance mode cancellation filter is limited to a predetermined level. 5. The apparatus of claim 8, wherein the predetermined level is -10 decibels to 3 decibels. The method of claim 2, wherein the angle correction unit The spatial inverse filtering apparatus considering the cognitive aspect of human, characterized in that for correcting the angle of the n-th speaker using the virtual loudspeaker positioning method from the corrected result. The method of claim 10, wherein the angle correction unit A first virtual sound source synthesizer that considers a human directional psychological factor and calculates high and low frequency components separated from the corrected result and gains determined using the angle of the nth speaker; And And a tone correction filter configured to correct an angle of the n-th speaker by a difference between HRTFs at an actual position and a virtual position of the n-th speaker using the result calculated by the first virtual sound source synthesis unit. Spatial inverse filtering device considering cognitive aspect. The method of claim 11, wherein the first virtual sound source synthesis unit A component separator configured to separate the result corrected by the spatial transfer function correcting unit into the low frequency component and the high frequency component based on a predetermined reference frequency, and output the separated high frequency component and the low frequency component; A gain generator which considers the direction psychological factor of the human being and generates an energy gain and a speed gain using the determined energy and velocity vector directions and the angle of the nth speaker; A multiplier for multiplying the energy gain and the speed gain by the high frequency component and the low frequency component inputted from the component separator, and outputting the multiplied results; And And a first adder configured to add the multiplied results and output the added result to the tone correcting filter. The method of claim 11, wherein the tone correction filter A first HRTF calculating unit calculating an HRTF at the virtual position of the nth speaker; A second HRTF calculator configured to calculate an HRTF at an actual left position of the n-th speaker by using the measured position of the n-th speaker; A third HRTF calculator configured to calculate an HRTF at an actual right position of the n-th speaker by using the measured position of the n-th speaker; A second virtual sound source synthesizer for calculating high and low frequency components separated from gains determined using the angle of the n-th speaker and a result calculated by the second HRTF calculator; A third virtual sound source synthesis unit calculating the high and low frequency components separated from the gains determined using the angle of the n-th speaker and the result calculated by the third HRTF calculator, in consideration of the human directional psychological factor; A second adder configured to add the results calculated by the second and third virtual sound source synthesizers; A second octave band smoothing unit for octave band smoothing the result calculated by the first HRTF calculating unit; A third octave band smoothing unit for octave band smoothing the result added by the second adding unit; A component difference calculator for calculating a difference between output components of the second and third octave band smoothing units; And The component calculated by the inverse of the difference calculated by the component difference calculator, the result calculated by the first virtual sound source synthesis unit, and the filtered result as a result of correcting the angle of the n-th speaker Spatial inverse filtering device in consideration of the cognitive aspect of human, characterized in that it comprises a magnitude inversion unit for outputting. The method of claim 13, wherein the second or third virtual sound source synthesis unit A component which separates the result calculated by the second or third HRTF calculation unit into the low frequency component and the high frequency component based on a predetermined reference frequency, and outputs the separated high frequency component and the low frequency component Separator; A gain generator which considers the direction psychological factor of the human being and generates an energy gain and a speed gain using the determined energy and velocity vector directions and the angle of the nth speaker; A multiplier that multiplies the energy gain and the speed gain by the high frequency component and the low frequency component input from the component separator, and outputs multiplied results; And And a third adder configured to add the multiplied results and output the added result to the tone correction filter. 15. The method of claim 12 or 14, wherein the gain generator And generating the energy gain and the speed gain as follows.

(Where g _nE represents the energy gain, g _nV represents the speed gain, θ ₁ represents the angle between the position of the measurement microphone and the n th speaker positioned at the position where the user is expected to be located, θ ₂ represents the angle between the measurement microphone positioned at the position where the user is expected to be located and the n-th speaker and the neighboring speaker, and θ _V ^target and θ _E ^target are the θ ₁ indicated by the desired sound image direction and the measurement result. And a target value in the direction of the velocity vector and a target value in the direction of the energy vector are respectively set by using and θ _2. ) 15. The method of claim 12 or 14, wherein the gain generator And generating the energy gain and the speed gain as follows.

(Where g _nE represents the energy gain, g _nV represents the speed gain, θ ₁ represents the angle between the position of the measurement microphone and the n th speaker positioned at the position where the user is expected to be located, θ ₂ represents the angle between the measurement microphone positioned at the position where the user is expected to be located and the n-th speaker and the neighboring speaker, and θ _V ^target and θ _E ^target are the θ ₁ indicated by the desired sound image direction and the measurement result. And a target value in the direction of the velocity vector and a target value in the direction of the energy vector are respectively set by using and θ _2. ) 15. The apparatus of claim 12 or 14, wherein the reference frequency is previously determined as a frequency dividing the directional psychological factor in a frequency band. 18. The apparatus of claim 17, wherein the reference frequency is 700 kHz. In the spatial inverse filtering method for filtering the first to N-th (where N is a positive integer greater than or equal to 2) audio output signals to the first to N-th speakers, respectively, Filtering the first to Nth audio input signals, respectively; And Extracting N filtered results by combining 2N filtered results, Filtering the n (1 ≦ n ≦ N) audio input signals Measuring a spatial transfer function of the nth speaker and an angle of the nth speaker from an nth audio input signal; And Correcting the spatial transfer function of the n-th speaker and correcting the angle of the n-th speaker by using the corrected result. A first to Nth, where N is a positive integer of 2 or more. A computer program storing a computer program for controlling a spatial inverse filtering device for filtering audio input signals and outputting the first to Nth speakers, respectively. In the recording medium, the computer program is Filtering the first to Nth audio input signals, respectively; And Extracting N filtered results by combining 2N filtered results, Filtering the n (1 ≦ n ≦ N) audio input signals Measuring a spatial transfer function of the nth speaker and an angle of the nth speaker from an nth audio input signal; And And correcting the spatial transfer function of the n-th speaker, and correcting the angle of the n-th speaker using the corrected result.

Images