JPH01233498A - speech encoding device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] 〔overview〕 Regarding a speech encoding device used for high-efficiency encoding of speech, etc.

The purpose of this invention is to reduce the transmission bit rate by using a pitch cutting method and a band limiting method, while suppressing the disadvantages of these methods, such as increased encoding delay and distortion of reproduced sound.

Pitch detection means for detecting the pitch period of an audio signal; and a pitch waveform that samples the audio waveform for a plurality of pitches based on the pitch period detected by the pitch detection means and generates a waveform for one pitch from the waveform for the plurality of pitches. The apparatus includes a generating means, a band limiting means for band-limiting the frequency band of the 1-pitch waveform generated by the pitch waveform generating means, and an encoding means for encoding the audio waveform band-limited by the band limiting means.

The number of waveforms for a plurality of pitches in the pitch waveform generating means and the limited bandwidth by the band limiting means are changed in accordance with the size of the pitch period detected by the pitch detecting means.

[Industrial application field]

The present invention relates to a speech encoding device used for high-efficiency encoding of speech.

An audio encoding device is required to be able to encode an audio signal at a low bit rate while reproducing the original audio without causing any audible discomfort on the playback side.

[Conventional technology]

One method of high-efficiency encoding is to sample N-pitch waveforms from an audio signal and create a 1-pitch audio waveform from these N-pitch audio waveforms.

A pitch extraction method is known in which the received signal is encoded and transmitted to the receiving side, and the receiving side decodes the received signal and repeats it N times to reproduce the original N pitches worth of audio signal.

According to this method, the transmission bit rate can be reduced to 1/N compared to the case where all audio waveforms are transmitted.

As another method of high-efficiency encoding, a method is also known in which the sampling frequency is reduced by band-limiting the audio signal, thereby lowering the bit rate.

In other words, the band of the audio signal is reduced to 1/M by 1lffl, and 1/
By downsampling at a sampling frequency of M, the transmission bit rate is reduced to 1/M compared to the case where no band limitation is performed.

[Problem that the invention seeks to solve]

The former pitch extraction method, which generates a one-pitch waveform from a plurality of pitch waveforms, has a problem in that the pitch becomes too large at bass frequencies, probably due to encoding delay.

That is, in the encoding delay on the transmitting side, the pitch period is T,
If the number of sample waveforms of the multi-pitch waveform from which the 1-pitch waveform is cut out is N, then generally τ=2N-T. Now set the maximum pitch period Tmax to 2Qmse
c.

When the number N of sample waveforms is 6, the maximum encoding delay τw
The ax is 240 m5ec, and this size poses a practical problem for telephone calls. Therefore, the size of the number N of sample waveforms is limited by the maximum pitch period, and therefore a sufficiently low bit rate cannot be achieved.

The latter method of band-limiting the audio signal has a problem in that when the band-limited audio signal is reproduced on the receiving side, the sound becomes muffled to the auditory sense.

Therefore, an object of the present invention is to provide an audio encoding device that reduces the transmission bit rate by using a pitch cutting method and a band limiting method, while suppressing the disadvantages of these methods, such as an increase in encoding delay and muffled reproduced sound. Our goal is to provide the following.

[Means for solving problems]

FIG. 1 is a block diagram illustrating the principle of a speech encoding device according to the present invention.

The speech encoding device according to the present invention includes a pitch detection means 1 for detecting the pitch period T of the speech signal, and a plurality of pitches of the speech signal sampled based on the pitch period T detected by the pitch detection means 1 and N pitches. The pitch waveform generating means 2 generates a waveform for one pitch from the audio waveform of , and the frequency band of the one pitch waveform generated by the pitch waveform generating means 2 is
It is equipped with a band limiting means 3 that limits the band to M, and an encoding means 4 that encodes the audio waveform whose band has been band limited by the band limiting means 3. The number of samples N of the pitch waveform in the pitch waveform generating means 2 and the rate M of band limiting by the band limiting means 3 are changed accordingly.

[Effect]

Normally, the pitch period of one person's voice is 80)1z or more, and it may sometimes be lower than this depending on intonation. Therefore, speech with a long pitch period T, which causes problems due to encoding delay, will mostly appear when the intonation is low, but for such speech with low intonation, the frequency band should be limited on the transmitting side. However, the playback audio on the receiving side does not sound muffled to the auditory sense.

Bandwidth limitations have little effect in practical terms.

Therefore, using these auditory characteristics, the coded via)
To perform audio encoding that reduces the rate, shortens the encoding delay, and eliminates the distortion of reproduced sound. In other words, for speech with a long pitch period T, the pitch waveform generating means 1 reduces the number of pitch waveform samples N to increase the encoding delay τ (bits due to the reduction in the number of pitch waveform samples N) To increase the rate, limit the audio waveform band to 1/M using the band limiting means 3 to reduce the bit rate to 1/M.
offset by reducing the Even if the band is limited in this manner, since the voice has a long pitch period, the effect of the band limit is not very perceptible on the playback side.

For audio with a short pitch period T, the pitch waveform generation signal 2 increases the number of pitch waveform samples N to reduce the bit rate, and the band restriction signal 3 relaxes the degree of band restriction to make the reproduced audio muffled. This prevents the sound from becoming too loud.

In this way, in the present invention, the number N of pitch waveform samples and the band limiting rate 1/M are controlled according to the pitch period T.
When T is large, the pitch waveform sample N is made small to shorten the encoding delay τ.

Instead, by increasing M, the encoding compression rate is 1/
While keeping L=1/NM almost constant.

The quality of the reproduced sound is aurally equivalent to when no band limitation is applied.

For example, when the pitch period T=0 to 12.5 m5ec, the number of samples is N-6 and the band limit rate is 1/M=1.

On the other hand, when the pitch period T = 12.5 to 20 m5ec, the number of samples N and the band restriction rate 1/M are set according to the pitch period T so that the number of samples N-3 and the band restriction rate 1-/M = 1/2. In the former case, the maximum value of encoding delay r wax is 2 x 12.5 x 6 = 150
msec + maximum encoding delay τl1ax in the latter case
is 2 x 20 x 3 = 120 m5ec, and the encoding delay is about 150 m5ec at maximum, which can be set to a level that does not pose a problem in practice.

〔Example〕

The present invention will be described in detail below with reference to nine drawings.

An encoding unit according to an embodiment of the present invention is shown in FIG. In FIG. 2, the audio signal S is input to a pitch extraction circuit 10 and a 1/N extraction circuit 11.

The pitch extraction circuit 10 is a circuit that extracts the pitch period of the input speech waveform, and the extracted pitch period T is sent to the 1/N extraction circuit 11 and the switching circuit 15, and is also transmitted to the decoding section via the transmission line. Ru.

The 1/N extraction circuit 11 is a circuit that generates a 1-pitch audio waveform from an N-pitch input audio waveform.

The pitch period T extracted by the pitch extraction circuit 10 is 15 m.
In the case of 5ec or more, create a 1-pitch waveform from N=3, that is, 3-pitch audio waveform, and pitch period T <
In the case of 15m5ec, N=6, that is, a 1-pitch audio waveform is created from 6 pitches' worth of audio waveforms.

The 1-pitch waveform generated by the 1/N extraction circuit 11 is then input to the band division filter 12. The band division filter 12 divides the input audio signal S with a bandwidth of O~4kHz into O~2
The low frequency signal St of kHz and the high frequency signal SH of 2 to 4 kHz are divided and sent to encoders 13 and 14 for encoding, and these low frequency signal SL and high frequency signal SL1 is the sampling signal of the original audio signal.
is downsampled to.

The low frequency signal SL from the encoder 13 is sent as is to the transmission path, and the high frequency signal SH from the encoder 14 is sent to the switching circuit 15.
It is sent out to the transmission path via. The switching circuit 15 receives pitch period T information from the pitch extraction circuit 10 + T<15
When m5ec, it is closed and sends out the high frequency signal So of the encoder 14 to the transmission path, while when T≧15m5ec, it is opened and cuts off the sending of the high frequency signal SH of the encoder 14 to the transmission path. It is composed of

In this way, this embodiment uses a band division coding method as a band limiting method in the encoding section, that is, a method in which human power is divided into high frequency components and low frequency components and the signals of each band are independently encoded. The signal of this special number band is downsampled according to its bandwidth. .

A decoding section of an embodiment according to the present invention is shown in FIG. In FIG. 3, a low frequency signal SL sent from the encoding section via a transmission path is input to a decoder 20, and a high frequency signal 3M is input to a decoder 21 via a switch 24.

Further, the pitch period T information is input to the switch 24 and the N-times repeat circuit 23.

The switch 24 is a circuit that is switched according to the pitch period T, and when T<15m5ec, it is switched to the transmission line side and inputs the high frequency signal SR from the transmission line to the decoder 21, and when T≧15m5ec, it is switched to the transmission line side and inputs the high frequency signal SR from the transmission line to the decoder 21. It is configured to cut off input of the high frequency signal SH to the decoder 2I.

The output signals of the decoders 20 and 2I are input to a band synthesis filter 22 and combined, and the combined signal is input to an N-times repeating circuit 23. The N-times repeating circuit 23 is a circuit that repeats the decoded audio waveform from the band synthesis filter 22 N times based on the pitch period T to create reproduced audio.

The operation of the example system is described below. First, in the encoding section, the input audio signal S is input to the pitch extraction circuit 10 and the 1/N extraction circuit 11.

A pitch extraction circuit 10 extracts the pitch period T of the audio signal S. Now, this extracted pitch period T is T<15m
Suppose that it is 5ec. Based on this pitch period T, the 1/N extraction circuit 11 samples the input audio signal by 6 pitches and generates and outputs a 1-pitch audio waveform from the 6-pitch waveform since T<15m5ec. The 1 pitch audio waveform from the 1/N extraction circuit 11 is input to the band division filter 12 where it is divided into a low frequency signal SL and a high frequency signal SH. 14. The switch 15 is closed because the pitch period T<15m5ec.
Therefore, both the low frequency signal SL and the high frequency signal So from encoders 13 and 14 are transmitted to the decoding section via the transmission path.

On the other hand, when the pitch period T extracted by the pitch extraction circuit 10 is T≧15m5ec, the 1/N extraction circuit 11 samples the audio signal S for 3 pitches and uses the audio waveform for 3 pitches to extract 1 pitch from the audio waveform for 3 pitches. Generates audio waveform. This audio waveform is divided into a low frequency signal SL and a high frequency signal SH by the band division filter 12 and encoded by encoders 13 and 14, as described above.

Since the switch 15 is open when T≧15m5ec.

The high frequency signal So from the encoder 14 is not sent out to the transmission path.

In this way, when the pitch period T is T≧15m5ec, 1
/N The number of pitch waveform samples N in the extraction circuit 11 is T<
Since it is half of that of 15m5ec, the encoding compression rate in this 1/N extraction circuit 11 is reduced by half, but the low frequency signal S divided by the band division filter 12 of the audio signal S is
Since only t is sent to the decoding unit, the bit rate can be halved, and therefore the encoding compression rate of the signal sent to the transmission path is ultimately .

The pitch period T is the same as when T<15 m5ec.

In other words, if the number of pitch waveform samples is N, and the band is limited to 1/M and downsampled to 1/M, then the compression ratio 1/L = 1/ (N - M) is always constant regardless of the pitch period T. constant.

In the decoding section, when T<15m5ec, the switch 24 is connected to the transmission line side, and therefore the low frequency signal St and the high frequency signal SH transmitted via the transmission channel are input to the decoders 20 and 21. decrypted.

Thereafter, the signals are synthesized by the band synthesis filter 22 and the synthesized signal is inputted to the repeating circuit 23 N times. The N-time repeat circuit 23 repeats the waveform of this composite signal six times to generate a reproduced signal.

When T≧15 m sec, only the low-frequency signal St from the transmission path is decoded by the decoder 20 and inputted to the N-times repeating circuit 23 via the band synthesis filter 22, and the N-times repeating circuit 2
3, the synthesized signal waveform is repeated three times to generate a reproduced signal.

Various modifications are possible in implementing the invention. In the above-described embodiment, a band division encoding method was used as a band limiting method in the encoding section.

Of course, the invention is not limited to this.9 For example, discrete Fourier transform (DFT)
You can also use That is, by performing discrete Fourier transform on the input audio signal to extract a line spectrum, and removing high frequency components of this line spectrum according to the pitch period T, band uniformity can be performed. In this case as well, the high frequency components of the line spectrum are reduced so that the compression ratio remains approximately constant.

〔Effect of the invention〕

According to the present invention, it is possible to perform audio encoding at a low bit rate while preventing the encoding delay from becoming too large or the reproduced sound becoming muffled on the playback side when encoding the audio signal. can.

[Brief explanation of the drawing]

FIG. 1 is a diagram explaining the principle of the present invention. FIG. 2 is a block diagram of an encoding unit according to an embodiment of the present invention;
and FIG. 3 is a block diagram of a decoding section of an embodiment according to the present invention. In fig. 1 - Pitch detection means 2 - Pitch waveform generation means 3 - Band limiting means 4 - Encoding means IO - Pitch extraction circuit 11 - 1/N extraction circuit 12 - Band division filters 13, 14 - - Encoder 15, 24 - Switcher 20, 21 - Decoder 22 - Band synthesis filter 23 - N-times repeat circuit 41 4-way dir Sun Moon no Ura Sai Samurai 1] Tf number Igoya Figure 2

Claims (1)

[Claims] 1. Pitch detection means for detecting the pitch period of an audio signal (
1), and pitch waveform generating means (2) which samples the audio signal for a plurality of pitches based on the pitch cycle detected by the pitch detection means (1) and generates a waveform for one pitch from the audio waveform for the plurality of pitches. ), band limiting means (3) for band limiting the frequency band of the one pitch waveform generated by the pitch waveform generating means (2), and encoding the audio waveform band limited by the band limiting means (3). the number of waveform samples in the pitch waveform generating means (2) and the band limiting means according to the size of the pitch period detected by the pitch detecting means (1); (3) A speech encoding device configured to change the limited bandwidth according to (3).