KR100195707B1 - A digital audio signal converter - Google Patents

Abstract

The present invention relates to a digital audio signal conversion apparatus, comprising: an AC-3 audio decoder (100) for restoring and outputting an encoded AC-3 audio signal to a PCM audio signal; A sampling frequency and bit rate control unit (200) for converting the sampling frequency and bit rate of the PCM audio signal into a sampling frequency and bit rate applicable to an MPEG audio coding scheme and outputting the converted sampling frequency and bit rate; MPEG audio coder 300 for MPEG audio encoding and outputting the PCM audio signal output from the sampling frequency and bit rate control unit 200, MPEG audio to the digital audio signal compressed by the AC-3 audio coder It can be converted to restore using a decoder.

Description

Digital Audio Signal Inverter

1 is a block diagram of a typical MPEG audio coder.

2 is a block diagram of a typical MPEG audio decoder.

3 is a block diagram of a typical AC-3 audio encoder.

4 is a block diagram of a typical AC-3 audio decoder.

5 is a block diagram of a digital audio signal conversion apparatus according to the present invention.

* Explanation of symbols for main parts of the drawings

2: analysis subband filter 4: psychoacoustic model

6 bit allocation and quantizer 8 bit stream array

10: bit allocation decoder 12: dequantizer

14 scale factor decoder 16 synthesized subband filter

20: analysis filter bank 22: spectral envelope encoder

24 bit allocation unit 26 quantizer

28: frame arrangement unit 30: frame inverse arrangement unit

32: bit allocation unit 34: dequantizer

36: spectra envelope encoder 38: synthesis filter bank

100: AC-3 audio decoder

200: sampling frequency and bit rate control unit

210: input buffer 220: parameter detection unit

230: parameter converter 300: MPEG audio encoder

The present invention relates to a digital audio signal converter, and more particularly, to a digital audio signal converter that converts a digital audio signal compressed by an AC-3 audio encoder to be recovered using an MPEG audio decoder.

Digital audio became the standard for audio equipment in the 80s with the development of mass storage media such as CDs and DATs. However, since digital audio data has a large amount of information, it is essential to compress audio data for use in a medium having a limited bandwidth such as over-the-air broadcasting.

Therefore, various high quality audio compression technologies have been developed since the late 80's, and these technologies are commonly combined with existing data compression companies in consideration of human hearing characteristics.

Representative examples of the audio compression technology include the MPEG method and the AC-3 method limited by Dolby Laboratories.

In other words, MPEG method is defined by MPEG / ISO which decides standard of video and audio compression method that is added. MPEG-1 is a coding method that can compress video and audio at about 1.5 Mbit / s. -pattern adapted Universal Subband Integrated Coding And Multiplexing) is used, and MPEG-1 has been extended to MPEG-2 of multi-channel structure with a data rate of 6Mbit / s or more for digital broadcasting.

The MUSICAM method is a subband coding method using auditory characteristics, so that a reconstructed sound identical to the original sound can be obtained subjectively from 96 to 128 Kbit / s.

AC-3 is a standard for HDTV audio compression technology in North America. AC-3 uses an adaptive transform coding method using the Modified Discrete Cosine Transform (MDCT). From 2 to AC-2a, it was developed into AC-3 having a multichannel structure.

The MPEG audio coder using the MUSICAM scheme includes an analysis subband filter (2) for converting and outputting an input signal (PCM data) into a plurality of subband samples as shown in FIG. Fast Fourier Transform (FFT) of the input signal to obtain spectral information, obtain a masking threshold from the spectral information, and then obtain the difference between the masking threshold and the sound pressure level of each subband sample determined from the spectral information. A psychoacoustic motel 4 which calculates and outputs a signal to mask ratio (SMR); Bit allocation and quantizer 6 for allocating bits to the respective subband samples using the signal-to-mask ratio output from the psychoacoustic motel 4 and quantizing and outputting the respective subband samples according to the allocated bits. ); A bit stream array unit 8 that formats and outputs subband samples, additional information such as bit allocation information, and scale information, which are quantized and output by the bit assignment and quantizer 6, into a bit stream It is configured to include.

In the MPEG audio encoder configured as described above, the analysis subband filter 2 stores 32 new audio samples input in a buffer having a size of 512 samples in turn, and multiplies the buffer by an analysis window. We divide the 512 samples into eight 64-sample blocks and add each block to form a new vector.

This is multiplied by an analysis matrix such as the following equation 1 to make 32 subband samples.

The psychoacoustic model 4 then determines the maximum noise level that is masked by the original sound in each subband and uses the noise level (masking threshold) to determine the bit allocation that determines the actual quantizer of each band. can do.

At this time, the MPEG method provides two psychoacoustic models. The psychoacoustic motel 1 distinguishes pure and noise components from a spectrum of a signal, calculates individual masking thresholds of pure and noise, and an absolute audible limit. In consideration of the above, the signal-mask ratio is obtained by calculating a total masking threshold and calculating a signal-to-mask ratio in each subband.

That is, the spectrum is obtained through a fast Fourier transform (FFT), and the sound pressure level in each subband is determined therefrom. Since the masking curve varies depending on whether the masking component is pure or noise, the pure and noise components must be found from the spectral information.

In other words, if the local maxima is greater than 7dB higher than the surrounding signal, it is regarded as pure sound and finds the pure sound component, and then obtains one noise component within one critical band from the rest of the spectrum.

The masking threshold obtained by the pure tone is obtained by applying the experimentally obtained masking function, and the total masking threshold is obtained as the sum of the individual masking threshold and the audible limit.

The signal-to-masking ratio (SMR) is calculated by calculating the difference between the sound pressure level in each subband and the masking threshold.

As a result, if the signal-to-mask ratio (SMR) is small, the sound pressure level of the signal is small or masking is high, so that the effective quantization can be performed with fewer bits.

In the psychoacoustic model 2, a process of calculating energy according to a critical band of a signal, a process of calculating a spreading function, a convolution, an excitation degree of the auditory nerve, a process of calculating a masking threshold in consideration of an absolute audible limit, and Then, the signal-to-mask ratio is obtained by calculating the signal-to-mask ratio in each subband.

The psychoacoustic model 2 convolves the fast Fourier transform spectrum with a spreading function, which is an excitation model of the auditory nerve, to obtain a masking threshold, thereby obtaining a large amount of computation but more accurate results.

The bit allocation and quantizer 6 then allocates bits for each subband sample output from the analysis subband filter 2 using the signal-to-mask ratio output from the psychoacoustic model 4. Each subband sample or the like is divided by a scale factor according to the allocated bits and quantized.

At this time, the scale factor calculation for each subband is performed before quantization, and the scale factor calculation is performed every 12 samples, and the maximum value of the absolute values of 12 samples is found to be 0 to 2. Normalize to

The bit stream arranging unit 8 arranges additional information such as quantized subband samples, allocation information, and scale information output from the bit allocation and quantizer 8 into MPEG frames and transmits them as bit streams. It is.

Meanwhile, the MPEG audio decoder which restores the compressed and transmitted bit stream as described above may implement the MPEG audio coder in reverse.

That is, the conventional MPEG audio decoder includes a bit allocation decoder 10 for decoding and outputting bit allocation information for each subband sample from the compressed and transmitted bit stream as shown in FIG. 2; An inverse quantizer (12) for inversely quantizing each subband sample of the bit stream according to the bit allocation information output from the bit allocation decoder (10); A scale factor decoder (14) for detecting scale factor information in the bit stream and then calculating and outputting each subband sample quantized by the inverse quantizer (12) using the scale factor as an original subband sample; Composed of a subband filter 16 for reconstructing an audio signal by synthesizing each subband sample output from the scale factor decoder 14.

In the MPEG audio decoder as described above, the bit allocation decoder 10 decodes and outputs bit allocation information for each subband sample from the compressed and transmitted bit stream, and the dequantizer 12 uses the bit allocation decoder ( The subband samples of the bit stream are inversely quantized and output according to the bit allocation information output from 10).

The scale factor decoder 14 detects scale factor information in the bit stream, and then calculates and outputs original subband samples for each subband sample quantized by the dequantizer 12 using the scale factor. The synthesized subband filter 16 synthesizes each subband sample output from the scale factor decoder 14 to restore and output an audio signal.

In this case, a matrix used for reconstructing each subband sample into an audio signal in the synthesis subband filter 16 is represented by the following equation.

On the other hand, the AC-3 audio encoder using the MDCT transform coding scheme, as shown in Figure 3, the analysis filter bank 20 for converting an input signal (PCM audio signal) to the frequency domain to output a frequency coefficient; A spectral envelope encoder 22 which outputs the exponential portion of the frequency coefficient by hatching a spectral envelope; A bit allocation section (24) for outputting bit allocation information for quantizing the mantissa section of the frequency coefficient according to the spectral envelope; A quantizer 26 for quantizing and outputting the mantissa of the frequency coefficient according to the bit allocation information; The spectral envelope and the quantized mantissa are arranged in an AC-3 frame, and are arranged in a frame arrangement unit 28.

In the AC-3 audio encoder as described above, the analysis filter bank 20 converts an input signal into a frequency domain, where each frequency coefficient is represented by an exponent part and a mantissa part.

The spectral envelope encoder 22 encodes the exponent portion into a spectral envelope, and outputs the spectral envelope to the bit allocation unit 24 and the frame arranging unit 30. The bit allocation unit 24 outputs the frequency according to the spectral envelope. The bit allocation information for quantizing the mantissa part is output to the quantizer 26.

The quantizer 26 quantizes the mantissa of the frequency coefficient according to the bit allocation information, and outputs the quantized part to the frame arranging unit 28. The frame arranging unit 28 AC-3 to the spectral envelope and the quantized mantissa. They are arranged in frames and output, and this AC-3 frame sequence becomes an AC-3 bit stream.

Meanwhile, the AC-3 audio decoder for restoring the audio signal compressed by the AC-3 audio encoder may implement the AC-3 audio encoder in reverse.

That is, the AC-3 audio decoder includes a frame inverse arrangement unit 30 for separating the spectral envelope and the mantissa of the encoded AC-3 bit stream, as shown in FIG. A bit allocation unit (32) for outputting bit allocation information according to the spectral envelope; An inverse quantizer (34) for inversely quantizing and outputting the mantissa according to the bit allocation information; A spectral envelope decoder (36) for decoding the spectral envelope and outputting an exponent portion; And a synthesis filter bank 38 for converting the exponent portion and the mantissa portion into the time domain and outputting the PCM audio signal.

In the AC-3 audio decoder as described above, the frame inverse array unit 30 separates the spectral envelope and the mantissa of the encoded AC-3 bit stream, respectively, so that the spectral envelope is a bit allocator 32 and a spectral envelope decoder. While outputting to 36, the mantissa part is output to the dequantizer 34.

In this case, the frame inverse array unit 30 must synchronize the encoded AC-3 bit stream and perform an error check.

The bit allocation unit 30 outputs bit allocation information to the inverse quantizer 34 according to the spectral envelope, and the inverse quantizer 34 inversely quantizes the mantissa unit according to the bit allocation information. Output to (30).

The spectral envelope decoder 36 decodes the encoded spectral envelope and outputs an exponent portion to the synthesis filter bank 38. The synthesis filter bank 38 converts the exponent portion and the mantissa portion into a time domain to convert the PCM audio signal. Will print

As described above, the conventional MPEG audio coder and the AC-3 audio coder are incompatible with each other because they have different compression methods.

That is, a signal encoded by the MPEG audio coder cannot be restored by the AC-3 audio decoder, and a signal encoded by the AC-3 audio coder cannot be restored by the MPEG audio decoder.

Accordingly, an object of the present invention is to solve a conventional problem as described above, and to provide a digital audio signal conversion apparatus for converting a signal encoded by an AC-3 audio coder to be recovered using an MPEG audio decoder. There is this.

The digital audio signal conversion signal according to the present invention for achieving the above object comprises: an AC-3 audio decoder for restoring the encoded AC-3 audio signal to a PCM audio signal; A sampling frequency and bit rate control unit for converting the sampling frequency and the bit rate of the PCM audio signal into a sampling frequency and a bit rate applicable to an MPEG encoding method and outputting the sampling frequency and bit rate; And an MPEG audio encoder for MPEG audio encoding and outputting the PCM audio signal output from the sampling frequency and bit rate control unit.

Therefore, the signal encoded by the AC-3 audio encoder can be recovered using the MPEG audio decoder.

Hereinafter, with reference to the accompanying drawings will be described in detail the present invention.

5 is a block diagram of a digital audio signal conversion device according to the present invention, wherein the digital audio signal conversion device according to the present invention restores and outputs an encoded AC-3 audio signal to a PCM audio signal. 100; A sampling frequency and bit rate control unit (200) for converting the sampling frequency and bit rate of the PCM audio signal into a sampling frequency and bit rate applicable to an MPEG audio coding scheme and outputting the converted sampling frequency and bit rate; And an MPEG audio encoder 300 for MPEG audio encoding and outputting the PCM audio signal output from the sampling frequency and bit rate control unit 200.

The AC-3 audio decoder 100 includes: a frame inverse array unit 30 for rearranging the encoded AC-3 audio signal to separate and output the spectral envelope, the mantissa unit, and the header information, respectively; A bit allocation unit (32) for outputting bit allocation information according to the spectral envelope; An inverse quantizer (34) for inversely quantizing and outputting the mantissa according to the bit allocation information; A spectral envelope decoder (36) for decoding the spectral envelope and outputting an exponent portion; And a synthesis filter bank 38 for converting the exponent part and the mantissa part into the time domain and outputting the PCM audio signal.

The sampling frequency and bit rate control unit 200 includes: an input buffer 210 for storing the PCM audio signal output from the AC-3 audio decoder 100; A parameter detector 220 which receives header information from the frame inverse array unit 30 of the AC-3 audio decoder 100 and detects and outputs a sampling frequency and a bit rate parameter; And a parameter converter 230 that adjusts the sampling frequency and bit rate of the PCM audio signal stored in the input buffer 210 to the sampling frequency and bit rate applicable to the MPEG audio coding method according to the sampling frequency and bit rate parameters. It is composed.

In addition, the parameter converter 230 determines whether the sampling frequency output from the parameter detector 220 is a sampling frequency applicable to the MPEG audio coding scheme, and if the sampling frequency is applicable to the MPEG audio coding scheme, the input buffer 210. PCM stored in the input buffer 210 at the sampling frequency closest to the sampling frequency applicable to the MPEG audio coding method without converting the sampling frequency of the PCM audio signal stored at The sampling frequency of the audio signal is adjusted, and if the bit rate detected by the parameter detector 220 is a bit rate applicable to the MPEG audio coding method, the bit rate applicable to the MPEG audio coding method is determined. Me to 210) Without converting the bit rate of the PCM audio signal, when the MPEG audio coding scheme bit rate that can not be applied to is adapted to adjust to the available bit rate applied to the MPEG audio encoding method.

In this case, the parameter converter 230 decreases the sampling frequency through decimation when the sampling frequency detected and output from the parameter detector 220 is greater than the sampling frequency applicable to the MPEG audio coding method, and sampling. If the frequency is smaller than the sampling frequency applicable to the MPEG audio coding method, the sampling frequency is increased through interpolation to match the sampling frequency of the MPEG audio coding method.

On the other hand, the MPEG audio encoder 300 includes: an analysis subband filter (2) for converting and outputting a PCM audio signal into a plurality of subband samples; Fast Fourier Transform (FFT) of the input signal to obtain spectral information, obtain a masking threshold from the spectral information, and then obtain the difference between the masking threshold and the music level of each subband sample determined from the spectral information. A psychoacoustic model 4 for calculating and outputting a signal to mask ratio (SMR); Bit allocation and quantizer 6 for allocating bits to the respective subband samples using the signal-to-mask ratio output from the psychoacoustic model 4 and quantizing and outputting the respective subband samples according to the allocated bits. ); A bitstream format unit 8 that formats and outputs subband samples, additional information such as bit allocation information, scale information, etc., which are quantized and output by the bit allocation and quantizer 6 into a bitstream; It is configured to include.

Referring to the operation and effects of the digital audio signal conversion apparatus according to the present invention configured as described above in detail as follows.

The AC-3 audio decoder 100 restores the encoded AC-3 audio signal to the PCM audio signal and outputs the converted AC-3 audio signal to the sampling frequency and bit rate control unit 200. The sampling frequency and bit rate control unit 200 generates the AC-3 audio signal. The sampling frequency and bit rate of the PCM audio signal output from the audio decoder 100 are converted into a sampling frequency and bit rate applicable to the MPEG audio encoding method, and output to the MPEG audio encoder 300. The MPEG audio encoder 300 is MPEG audio encoding is performed on the PCM audio signal output from the sampling frequency and bit rate controller 200.

That is, in AC-3, the frame inverse array unit 30 of the audio decoder 100 separates the spectral envelope, the mantissa unit, and the header information of the encoded AC-3 bit stream, respectively, and the spectral envelope is the bit allocation unit 32. And the output to the spectral envelope decoder 36, the mantissa to the inverse quantizer 34, and the header information to the parameter detector 210 of the sampling frequency and bit rate control unit 200.

In this case, the frame inverse array unit 30 must synchronize the encoded AC-3 bit stream and perform an error check.

The bit allocation unit 30 outputs bit allocation information to the inverse quantizer 34 according to the spectral envelope, and the inverse quantizer 34 dequantizes the mantissa unit according to the bit allocation information to perform a synthesis filter bank ( 30).

The spectral envelope decoder 36 decodes the encoded spectral envelope and outputs an exponent portion to the synthesis filter bank 38. The synthesis filter bank 38 converts the exponent portion and the mantissa portion into a time domain to convert the PCM audio signal. Will print

The PCM audio signal restored as described above is stored in the input buffer 210 of the sampling frequency and bit rate control unit 200, wherein the parameter detector 220 controls the frame inverse array unit 30 of the AC-3 audio decoder 100. The header information is received from the header information, and the sampling frequency and bit rate parameters are detected and output to the parameter converter 230, and the parameter converter 230 is stored in the input buffer 210 according to the sampling frequency and bit rate parameters. The sampling frequency and bit rate of the PCM audio signal are adjusted to the sampling frequency and bit rate applicable to the MPEG audio coding method.

That is, the parameter converter 230 determines whether the sampling frequency detected by the parameter detector 220 is a sampling frequency applicable to the MPEG audio coding scheme, and if the sampling frequency is applicable to the MPEG audio coding scheme, the input buffer 210. If the sampling frequency is not applicable to the MPEG audio coding method without converting the sampling frequency of the PCM audio signal stored in the PCM audio signal, the PCM audio stored in the input buffer 210 is the closest sampling frequency among the sampling frequencies applicable to the MPEG audio coding method. While adjusting the sampling frequency of the signal, it is determined whether the bit rate detected by the parameter detector 220 is a bit rate applicable to the MPEG audio coding method, and if the bit rate is applicable to the MPEG audio coding method, the input buffer 210 is applied. Of stored PCM audio signals Without converting the bit rate, MPEG bit rate is not applicable to audio coding methods and the available bit rate adjustment applies to the MPEG audio encoding method.

At this time, if the sampling frequency detected by the parameter detector 220 is larger than the sampling frequency applicable to the MPEG audio coding method, the sampling frequency is reduced by decimation, and the sampling frequency is applicable to the MPEG audio coding method. If smaller, the sampling frequency is increased through interpolation to match the sampling frequency of the MPEG audio encoding method.

As described above, the PCM audio signal output from the AC-3 audio decoder 100 is adjusted to the sampling frequency and bit rate at the sampling frequency and bit rate 200, and then input to the MPEG audio encoder 300.

In addition, the analysis subband filter 2 of the MPEG audio encoder 300 stores 32 input audio samples in a buffer having a size of 512 samples in turn, and multiplies the buffer by an analysis window. Divide the 512 samples into eight 64 sample blocks and add each block to form a new vector.

This is multiplied by an analysis matrix such as the following equation 1 to make 32 subband samples.

The psychoacoustic model 4 then determines the maximum noise level that is masked by the original sound in each subband and cannot be heard, and uses this noise level (masking threshold) to assign the bit to determine the actual quantizer of each band. can do.

In the MPEG method, two psychoacoustic models are provided. The psychoacoustic model 1 distinguishes pure and noise components from a spectrum of a signal, calculates individual masking threshold values of pure and noise, and generates an absolute audible signal. The signal-to-mask ratio is obtained by calculating the overall masking threshold value and calculating the signal-to-mask ratio in each subband.

That is, the spectrum is obtained through a fast Fourier transform (FFT), and the sound pressure level in each subband is determined therefrom. Since the masking curve varies depending on whether the masking component is pure or noise, the pure and noise components must be found from the spectral information.

That is, if the local maxima is greater than 7dB greater than the surrounding signal, the pure sound component is considered to be pure sound, and then one noise component is obtained within one critical band from the remaining spectrums.

The masking threshold obtained by pure noise is applied by applying the experimentally obtained masking function, and the total masking threshold is obtained by the sum of the individual masking threshold and the audible limit.

The signal-to-masking ratio (SMR) is calculated by calculating the difference between the sound pressure level in each subband and the masking threshold.

As a result, if the signal-to-mask ratio (SMR) is small, the sound pressure level of the signal is small or masking is high, so that the effective quantization can be performed with fewer bits.

In the psychoacoustic model 2, a process of calculating energy according to a critical band of a signal, a process of calculating a spreading function, a convolution, an excitation degree of the auditory nerve, a process of calculating a masking threshold in consideration of an absolute audible limit, and Then, the signal-to-mask ratio is obtained by calculating the signal-to-mask ratio in each subband.

The psychoacoustic model 2 convolves the fast Fourier transform spectrum with a spreading function, which is an excitation model of the auditory nerve, to obtain a masking threshold, thereby obtaining a large amount of computation but more accurate results.

The bit allocation and quantizer 6 then uses a signal-to-mask ratio output from the psychoacoustic model 4 to allocate a bit for each subband sample output from the analysis subband filter 2. It allocates and quantizes each subband sample normalized by dividing by a scale factor according to this allocated bit.

In this case, the scale factor calculation for each subband is performed before quantization, and the scale factor calculation is performed every 12 samples, and the maximum value of the absolute values of 12 samples is found to be 0 to 2. Normalize to

The bit stream arrangement unit 8 arranges and transmits additional information such as quantized subband samples, bit allocation information, and scale information output from the bit allocation and quantizer 8 into MPEG frames. .

As described above, according to the present invention, the digital audio signal compressed by the AC-3 audio encoder can be converted to be restored using an MPEG audio decoder.

Claims (5)

An AC-3 audio decoder 100 for restoring and outputting the encoded AC-3 audio signal to a PCM audio signal; A sampling frequency and bit rate controller (200) for converting the sampling frequency and bit rate of the PCM audio signal into a sampling frequency and bit rate applicable to an AC-3 audio coding scheme and outputting the converted sampling frequency and bit rate; And an MPEG audio encoder (300) for MPEG audio encoding and outputting the PCM audio signal output from the sampling frequency and bit rate control unit (200). The apparatus of claim 1, wherein the sampling frequency and bit rate control unit (200) comprises: an input buffer (210) for storing the PCM audio signal output from the AC-3 audio decoder (100); A parameter detector 220 which receives header information from the AC-3 audio decoder 100 and detects and outputs sampling frequency and bit rate parameters; And a parameter converter 230 that adjusts the sampling frequency and bit rate of the PCM audio signal stored in the input buffer 210 to the sampling frequency and bit rate applicable to the MPEG audio coding method according to the sampling frequency and bit rate parameters. Digital audio signal converter, characterized in that configured to. The method of claim 2, wherein the parameter converter 230 determines whether the sampling frequency output to the parameter detector 220 is a sampling frequency applicable to the MPEG audio coding method, and if the sampling frequency is applicable to the MPEG audio coding method. Without converting the sampling frequency of the PCM audio signal stored in the input buffer 210, if the sampling frequency is not applicable to the MPEG audio coding method, the input buffer is the closest sampling frequency among the sampling frequencies not applicable to the MPEG audio coding method. And a sampling frequency of the PCM audio signal stored in the 210 audio signal. The method of claim 2, wherein the parameter converter 230 determines whether the bat rate detected by the parameter detector 220 is a bit rate applicable to the MPEG audio coding method, and if the bit rate is applicable to the MPEG audio coding method. A digital audio signal characterized in that it is adjusted to a bit rate applicable to the MPEG audio coding method without converting the bit rate of the PCM audio signal stored in the input buffer 210 and not applicable to the MPEG audio coding method. Inverter. The method of claim 3, wherein the parameter converter 230 decreases the sampling frequency through decimation when the sampling frequency detected and output by the parameter detector 220 is greater than the sampling frequency applicable to the MPEG audio coding method. Sampling frequency The digital audio signal converting apparatus, characterized in that the sampling frequency is increased by interpolation if the sampling frequency is smaller than the sampling frequency applicable to the MPEG audio coding system to match the sampling frequency of the MPEG audio coding system.