JP3535008B2 - Audio processing method and apparatus - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice processing technique for encoding / decoding a voice signal, and particularly to an output rate of a voice coded information bit string output from a voice encoder and a voice buffer. Adjust the mismatch between the audio sample storage rate and the input rate of the audio coded information bit string input to the audio decoder and the output rate of the reproduced audio sample output from the audio buffer. Regarding the technology to adjust.

[0002]

2. Description of the Related Art Conventionally, in a voice communication apparatus, in order to reduce the communication data amount of a voice signal, a transmission side apparatus encodes a voice signal and transmits the encoded voice signal via a wireless or wired transmission path, and a reception side apparatus. Then, the coded audio signal is received and decoded. FIG. 15 (a) shows an example of the configuration of a conventional speech encoding device, and FIG. 15 (b) shows an example of the configuration of a conventional speech decoding device.

In this speech coding apparatus, for example, 8 kHz
The audio sample a1 sampled in 16 bits and quantized with 16 bits is input to the input audio buffer 11 in synchronization with the sampling clock k1 and temporarily stored therein, and the audio sample stored in the input audio buffer 11 is stored in one frame ( For example, every 160 ms, for example, 160 samples are input to the speech encoder 12 as b1. And
The voice encoder 12 encodes the input voice sample b1 at, for example, 2400 bps, and a frame synchronization signal (a signal indicating the start of one frame section) c input from an external device.
The audio encoded information bit string e1 is output at 48 bits / frame in synchronization with 1 and the bit clock d1. The frame synchronization signal c1 and the bit clock d1 are synchronized.

Further, in the speech decoding apparatus, a frame synchronization signal (a signal indicating the start of one frame section) f1 and a bit clock (for example, 2400b) input from an external apparatus.
ps) g1 in synchronization with the audio coded information bit string h1
(48 bits / frame) is input to the audio decoder 14, the audio decoder 14 decodes the audio encoded information bit string h1 input at 2400 bps, and the reproduced audio sample string i1 (for example, 160 samples / frame). Is output. The frame synchronization signal f1 and the bit clock g
1 is synchronized. Then, the reproduced voice sample string i1 is input to the output voice buffer 13 and the output voice buffer 1
The reproduced voice sample accumulated in 3 is output as j1 in synchronization with the sampling clock (for example, 8 kHz) l1.

FIG. 16 shows an input / output timing chart of the speech coding apparatus and speech decoding apparatus. FIG. 7A shows a frame synchronization signal (c1 or f1),
The interval of the signal level "L" indicating the synchronization timing is 20 ms in this example. FIG. 2B shows a bit clock (d1
Alternatively, g1), which is 2400 bps synchronized with the frame synchronization signal in this example. FIG. 7C shows a voice coded information bit string (e1 or h1). The voice coded information bit string is synchronized with a frame synchronization signal and a bit clock, and in this example, at a rate of 48 bits per frame (20 ms). Input and output. FIG. 7D shows the sampling clock (k1 or 11), which is 8 kHz in this example. FIG. 7E shows an input audio sample input in synchronization with the sampling clock or a reproduced audio sample (a1 or j1) output. In this example, input / output is performed at a rate of 160 bits per frame (20 ms). It

Here, the frame sync signal of FIG. 1A and the bit clock of FIG. 2B are synchronized, and the frame sync signal is valid once every 48 cycles of the bit clock ("
Since the sampling clock (e) in the figure is generated in the voice encoding device or the voice decoding device, it is not synchronized with the frame synchronization signal or the bit clock input from the external device.

[0007]

In the above speech encoder and speech decoder, if there is no error in the frequency of the bit clock input from the external device and it is always 2400 bps, the number of speech samples (160 samples / frame) And the number of bits of voice encoding information (48 bits / frame) can always be kept in a matched state as designed, and voice encoding processing and voice decoding processing can be normally performed. However, in order to improve the frequency accuracy of the bit clock input from the external device, the cost increases considerably, and generally, a certain amount of error (for example, ± 2.5%) is allowed in the frequency accuracy of the bit clock. Therefore, due to the frequency fluctuation of the bit clock, the voice encoding process and the voice decoding / reproducing process may not be normally performed.

For example, if the bit clock frequency is ± 2.
When it varies by 5%, the number of voice samples input / output per frame varies in the range of 160 ± 4 samples based on the number of encoded voice information bits (48 bits / frame). Therefore, the number of voice samples (160 samples / frame) and the number of voice coding information bits (48 bits / frame) cannot always be kept at the design values, and the normal number of input voice samples (160
Sample / frame), the number of input voice samples of one frame stored in the input voice buffer 11 is 160 ±
Since it fluctuates within the range of 4 samples, the voice decoding cannot be normally performed by the voice decoder 12, and
Also in the audio decoding process, the number of reproduced audio samples output from the audio decoder 14 and stored in the output audio buffer 13 should be 160 samples / frame.
Since the number of reproduced audio samples output to the output audio buffer 13 varies within the range of 160 ± 4 samples / frame, normal audio reproduction based on the reproduced audio samples cannot be performed.

For such a problem, it is conceivable that the number of audio samples stored in the input buffer 11 or the output buffer 13 is thinned out or repeated so as to adjust to a predetermined number of samples. Even if adjustment is made to a predetermined 160 samples by simple thinning out of audio samples or repetition of repetition, there is a problem in that continuity between audio samples deteriorates, and as a result, the quality of reproduced audio deteriorates. Note that the above problem can occur similarly when the bit clock is generated by using an oscillation source different from the sampling clock in the present device instead of giving the bit clock from the external device.

The present invention has been made in view of the above conventional circumstances, and adjusts the number of audio samples output from an audio buffer for performing audio encoding processing to a predetermined number without deteriorating the quality of reproduced audio. Then, it aims at providing the audio | voice processing method and apparatus which can implement | achieve the audio | voice encoding process normally. Further, according to the present invention, the number of reproduced audio samples output from the audio buffer for performing audio reproduction is adjusted to a predetermined number without deteriorating the quality of reproduced audio, and the audio decoding reproduction processing is normally realized. It is an object of the present invention to provide a voice processing method and device capable of performing the same. Another object of the present invention is to provide a voice communication device and a voice communication system for implementing the above method.

[0011]

In the audio processing method according to the present invention, an audio sample is accumulated in an audio buffer for each predetermined frame in synchronization with a sampling clock, and the accumulated audio sample is subjected to an audio code for each frame. In the audio encoding process of inputting to the coder and outputting as the audio encoded information bit string for each frame in synchronization with the bit clock, it is output from the audio buffer for performing the audio encoding process as follows. Adjust the number of audio samples to a predetermined number.

That is, when the audio samples for each frame accumulated in the audio buffer are over-sampled and the bit clock decreases and changes relative to the sampling clock (either one of the sampling clock and the bit clock or both of them). (Including the case where fluctuates)
Is performed by increasing the interval between the sample points and interpolating the voice sample at the new sample point from the original voice sample, so that the voice sample for each frame input to the voice coder is a voice that matches the bit clock. Convert to number of samples. Also, when the bit clock relatively increases and changes with respect to the sampling clock (same as above), by shortening the interval between sample points and interpolating the audio sample at the new sample point from the original audio sample,
The audio samples for each frame input to the audio encoder are converted into the number of audio samples that matches the bit clock.

Further, in the speech processing method according to the present invention, the inputted speech samples are stored in the speech buffer for each frame (P / frame), and the speech samples stored in the speech buffer are processed by the speech encoder. In the voice processing of performing voice encoding processing and outputting a voice encoded information bit string (Q bits / frame) in synchronization with a bit clock input from an external device or internally oscillated, The number of audio samples output from the audio buffer for performing the audio encoding process is adjusted to a predetermined number. In addition, P, Q, M, N, and L described in this specification are positive integers.

That is, the number of audio samples stored in the audio buffer is counted while the audio coded information bits are output as Q bits, and the count value becomes (P + (M *
N)), the audio samples stored in the audio buffer are oversampled L times (P +
(M * N)) * L after making L subdivided sample points,
For each (L + M) subdivided sample points, a new voice sample is interpolated from the adjacent original voice sample (L * N) times and output to the voice encoder.
By outputting (P- (L * N)) times voice samples to the voice encoder for each number, a total of P voice samples are output to the voice encoder, and the output rate from the voice encoder is output. To fit. Further, the count value is (P- (M *
N)), the audio samples stored in the audio buffer are oversampled L times (P−
(M * N)) * L after making L subdivided sample points,
For each (L−M) number of subdivided sample points, a new voice sample is interpolated (L * N) times from the adjacent original voice sample and output to the voice encoder.
By outputting (P- (L * M)) times voice samples to each voice encoder for each number, a total of P voice samples are output to the voice encoder and the output rate from the voice encoder is output. To fit.

By the voice encoding process as described above, for example, the bit clock fluctuates, and the accumulation rate of the voice samples stored in the voice buffer is higher than the rate of the encoded information bit string output from the voice encoder. If it is too small or too large, the voice samples are distributed without being biased in one frame, the voice samples are partially overlapped or deleted, and the number of voice samples output from the voice buffer is output from the voice encoder. The encoded information bit sequence is converted into a format adapted to the rate of the encoded information bit stream. Therefore, the normal voice coding process is performed, and the influence of the adjustment in the voice samples is finely dispersed, so that the continuity between the voice samples is maintained and the deterioration of the auditory sense is small. You can play various sounds.

Further, in the audio processing method according to the present invention, the audio encoding information bit string for each frame is input to the audio decoder in synchronization with the bit clock, the decoded reproduced audio sample is output, and each frame is processed. In the audio decoding / reproducing process of accumulating the reproduced audio samples in the audio buffer and outputting the reproduced audio samples for each predetermined frame in synchronization with the sampling clock, the audio is decoded and reproduced as follows. Therefore, the number of reproduced voice samples output from the voice buffer is adjusted to a predetermined number.

That is, when the reproduced audio samples for each frame accumulated in the audio buffer are over-sampled and the bit clock decreases and changes relative to the sampling clock (same as above), the interval between sample points is shortened. Then, by interpolating the audio sample at the new sample point from the original audio sample, the reproduced audio sample for each frame output from the audio buffer is converted into the number of audio samples that matches the sampling clock. Further, when the bit clock relatively increases and changes with respect to the sampling clock (same as above), the interval between sample points is lengthened and the audio sample at the new sample point is interpolated from the original audio sample, The reproduced audio samples for each frame output from the audio buffer are converted into the number of audio samples that matches the sampling clock.

Further, in the audio processing method according to the present invention, the audio coded information bit string (Q bits / frame) is input to the audio decoder for each frame in synchronization with the bit clock.
After the audio decoding processing is performed for each frame and the reproduced audio signal samples (P pieces / frame) are stored in the audio buffer, in the audio processing for outputting the reproduced audio sample for each frame, the audio decoding is performed as follows. The number of reproduced audio samples output from the audio buffer is adjusted to a predetermined number.

That is, the number of reproduced audio samples output from the audio buffer is counted while inputting Q bits of audio encoded information bits, and the count value is (P + (M
* N)), the P reproduced audio samples output from the audio decoder and stored in the audio buffer are oversampled L times to obtain (P * L) subdivided sample points. Then, for each (L−M) subdivided sample points, a new voice sample is interpolated from the original voice sample that is adjacent (L * N) times and output from the voice buffer. (P + (M * N)-(L
By outputting * N)) times audio samples from the audio buffer, a total of (P + (M * N)) reproduced audio samples are output to absorb fluctuations in the output rate from the audio decoder. When the count value is (P- (M * N)), the P reproduced audio samples output from the audio decoder and stored in the audio buffer are oversampled L times (P After making * L) subdivided sample points, interpolate a new audio sample from the original audio sample adjacent (L * N) times for each (L + M) subdivided sample points and output it from the audio buffer. , (P− (M * N) − (L * N)) times for each L subdivided sample points are output from the audio buffer,
A total of (P- (M * N)) reproduced speech samples are output to absorb fluctuations in the output rate from the speech decoder.

By the voice decoding reproduction process as described above, for example, the bit clock fluctuates, the rate of the encoded information bit string input to the voice decoder fluctuates, and the accumulation rate of the reproduced voice samples stored in the voice buffer changes. If it is too small or too large, the audio samples are distributed without being biased in one frame to partially overlap or delete the audio samples, and the reproduced audio samples output from the audio buffer are converted into a predetermined number. . Therefore, since the normal audio decoding / playback processing is performed and the influence of the adjustment in the playback audio sample is finely dispersed, the continuity between the playback audio samples is maintained and the auditory deterioration is small. High quality sound can be reproduced.

The voice processing apparatus according to the present invention comprises a voice buffer for storing the input voice samples, and a voice encoder for encoding the voice samples stored in the voice buffer, and the bit clock is used. In a speech processing device that outputs a speech coded information bit string for each frame from a speech encoder in synchronization, one frame of speech encoded information bits is stored in a speech buffer while being output from the speech encoder. An audio sample counter for counting the number of audio samples and a count value by the audio sample counter are equivalent to one frame to be stored in the audio buffer while one frame of audio encoded information bits is output from the audio encoder. An audio sample that converts the number of audio samples stored in the audio buffer when the number of audio samples differs from the specified number. And a, the number converter.

The audio sample number converter oversamples the audio samples for each frame stored in the audio buffer, and if the count value is greater than the predetermined number of samples, the interval between sample points is lengthened. By interpolating the voice sample at the new sample point from the original voice sample, the voice sample for each frame input to the voice encoder is converted into the predetermined number of samples, and the count value is less than the predetermined number of samples. In this case, by shortening the interval between sample points and interpolating the audio sample of the new sample point from the original audio sample,
The audio samples for each frame input to the audio encoder are converted into the predetermined number of audio samples. As a result, the above-described speech coding method is performed, the speech samples are distributed without being biased in one frame, the speech samples are partially overlapped or deleted, and the number of speech samples output from the speech buffer is determined by the speech encoder. It is converted to one that is compatible with the rate of the output audio coded information bit string to perform normal audio encoding processing, and the effects of adjustments in audio samples are finely dispersed, so High-quality sound can be reproduced while maintaining continuity between samples.

Further, the speech processing apparatus according to the present invention is a speech decoder for speech-decoding a speech coded information bit string for each frame, and a speech for accumulating and outputting reproduced speech signal samples subjected to speech decoding processing for each frame. And a voice processing device for inputting a voice coded information bit string to a voice decoder in synchronization with a bit clock, while the voice coded information bits for one frame are input to the voice decoder. The audio sample counter for counting the number of reproduced audio samples output from the audio sample counter and the count value by the audio sample counter are stored in the audio buffer while one frame of audio encoded information bits is input to the audio decoder. Number of audio samples stored in the audio buffer when it differs from the predetermined number of reproduced audio samples for one frame And a, a number of audio samples converter for converting.

The audio sample number converter oversamples the reproduced audio samples for each frame stored in the audio buffer and lengthens the interval between sample points when the count value is larger than the predetermined number of samples. By interpolating the audio sample at the new sample point from the original audio sample, the reproduced audio sample for each frame output from the audio buffer is converted into the predetermined number of samples, and the count value is less than the predetermined number of samples. In this case, the interval between sample points is shortened and the audio sample at the new sample point is interpolated from the original audio sample to convert the reproduced audio sample for each frame output from the audio buffer into the predetermined number of audio samples. To do.
As a result, the above method of voice decoding and reproduction is carried out,
The audio samples are distributed without being biased in one frame and the audio samples are partially overlapped or deleted, and the reproduced audio samples output from the audio buffer are converted into a predetermined number to perform normal audio decoding and reproduction processing. Since the influence of the adjustment in the reproduced voice sample is finely dispersed, it is possible to maintain the continuity between the reproduced voice samples and reproduce the high quality voice.

Further, a voice communication device according to the present invention is a voice communication device which encodes and transmits a voice signal and decodes a received encoded voice signal. High-quality voice communication is realized by providing the above-described voice processing device for performing voice decoding and reproduction in the signal encoding processing unit and for providing the voice signal decoding processing unit. Further, the voice communication system according to the present invention performs voice coding in the voice communication system in which the transmitting side device encodes and transmits a voice signal and the receiving side device decodes the received encoded voice signal. A voice processing device is provided in the voice signal encoding processing unit of the transmission side device, and the voice processing device described above for performing voice decoding and reproduction is provided in the voice signal decoding processing unit of the reception side device to realize high-quality voice communication. .

[0026]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be specifically described based on Examples. FIG. 1 shows the configuration of a speech coding apparatus according to an embodiment of the present invention, and FIG. 2 shows the configuration of a speech coding apparatus according to an embodiment of the present invention. In a voice communication device such as a wireless communication terminal device that encodes and transmits a voice signal and decodes a received encoded voice signal, the voice encoding device according to the present embodiment is applied to a voice signal encoding processing unit. In addition, the speech decoding apparatus of this embodiment is applied to a speech signal decoding processing unit. Further, this voice communication device constitutes a voice communication system in which a voice communication device on the transmission side encodes and transmits a voice signal and a voice communication device on the receiving side decodes the encoded voice signal received.

First, in the speech coder shown in FIG. 1, the input speech sample a2 sampled at 8 kHz and quantized with 16 bits is used as the sampling clock k.
2 is input to the input audio buffer 1 and is temporarily stored in synchronism with 2, and the audio samples stored in the input audio buffer 1 are, for example, 16 for each frame (for example, 20 ms).
It is input to the voice encoder 2 as b2 every 0 samples.
Then, the audio encoder 2 encodes the input audio sample b2 at, for example, 2400 bps, and synchronizes the audio encoded information bit string e2 with, for example, 48 bits in synchronization with the frame synchronization signal c2 and the bit clock d2 input from the external device. / Output frame. Note that, as in the case shown in FIG. 16, the frame synchronization signal c2 and the bit clock d2 are synchronized, and the sampling clock k2 is generated in the audio encoding device. It is not synchronized with the bit clock d2.

In this speech coding apparatus, the speech sample counter 3 obtains the number of speech samples a2 input in one cycle of the frame synchronization signal c2 by counting the sampling clock k2, and the speech samples per frame. The number m2 is output to the voice sample converter 4. This voice sample number converter 4 has a voice sample number m.
2 is converted into a predetermined number of voice samples (160 samples in this example). That is, the audio sample number converter 4 once reads out the audio sample sequence n2 stored in the input audio buffer 1, converts it into a predetermined number of audio samples, and then converts the converted audio sample sequence o2 into the input audio buffer 1.
, And re-input, and the voice sample sequence converted into the predetermined number is output to the voice encoder 2.

In the speech decoding apparatus shown in FIG. 2, the speech coded information bit string h2 (for example, 48 bits / bit) is synchronized with the frame synchronization signal f2 and the bit clock (for example, 2400 bps) g2 input from the external apparatus. Frame) to the audio decoder 6, and the audio decoder 6 inputs the audio encoded information bit string h input at 2400 bps, for example.
2 is decoded, and the reproduced voice sample sequence i2 (for example,
160 samples / frame) to the output audio buffer 5. Then, the reproduced voice sample sequence i2 is temporarily stored in the output voice buffer 5, and the stored reproduced voice samples are sampled at a sampling clock (for example, 8 kHz) l.
It is output to an audio reproducing device (not shown) as j2 in synchronization with 2. As shown in FIG. 16, since the frame synchronization signal f2 and the bit clock g2 are synchronized and the sampling clock 12 is generated in the audio decoding device, the frame synchronization signal f2 and bits input from the external device are input. It is not synchronized with the clock g2. In this example, the bit clocks d2 and g2 are input from an external device, but the bit clock is the sampling clock k2 in the device.
Even when it is given from an oscillation source different from 12, the same effects can be obtained by the present invention.

In this audio decoding device, the audio sample counter 7 obtains the number of reproduced audio samples output from the output audio buffer 5 in one cycle of the frame synchronization signal f2 by counting the sampling clock l2, and one frame is obtained. The number p2 of reproduced voice samples output per hit is output to the voice sample number converter 8. The audio sample number converter 8 converts the reproduced audio sample number p2 stored in the output audio buffer 5 into a predetermined reproduced audio sample number. That is, the audio sample number converter 8 once reads the reproduced audio sample sequence q2 stored in the output audio buffer 5, converts it to a predetermined reproduced audio sample number, and then converts the converted reproduced audio sample sequence r2 into the output audio buffer 5 , And re-input, and the reproduced audio sample sequence converted into the predetermined number is output to an audio reproducing device (not shown).

Next, the processing by the above speech coding apparatus will be described with reference to FIGS. In this audio encoding device, a digital data output interrupt process for outputting the audio encoded information bits e2 from the audio encoder 2 and converting the number of audio samples stored in the input audio buffer 1 is performed by the processing procedure shown in FIG. Also, the voice sample input interrupt process for inputting the voice sample a2 to the input voice buffer 1 is performed by the processing procedure shown in FIG.

First, the digital data output interrupt process shown in FIG. 3 is generated at each falling edge of the bit clock d2, and the digital data (voice coded information bit e2) is output from the voice coder 2 by one bit (step). S
1). Then, the level of the frame synchronization signal c2 is confirmed (step S2), and if the level is not "L", the process returns (ends the interrupt routine) while the level is "L".
If it is L ", it is a delimiter of frame processing, and therefore the following processing is continued in order to adjust the number of audio samples output from the input audio buffer 1 as necessary.

That is, the count value of the voice sample counter 3 is confirmed (step S3), and when the count value is a predetermined value P (160 samples in this example) that matches the output rate from the voice encoder 2. Since it is not necessary to adjust the number of voice samples, the count value of the voice sample counter 3 is cleared (step S7) and the process returns. On the other hand, when the count value is not the predetermined value P, it does not match the output rate from the audio encoder 2 at the present time due to the fluctuation of the bit clock d2, and therefore the input audio buffer 1
It is determined whether or not the count value is larger than the predetermined value P in order to adjust the number of audio samples output from (step S
4).

As a result, if the count value is smaller than the predetermined value P (P-M * N in this example), the voice sample number increasing process described later is performed (step S5), and the count value is predetermined. Greater than P (in this example, P + M * N
In this case, the number of audio samples is reduced as will be described later (step S6), the number of audio samples of one frame accumulated in the input audio buffer 1 is adjusted to a predetermined number P, and the P audio The sample sequence b2 is output from the input voice buffer 1 to the voice encoder 2. Then, the count value of the audio sample counter 3 is cleared for the next frame processing (step S7), and the process returns.

The voice sample input interrupt process shown in FIG. 4 is generated at each falling edge of the sampling clock k2, and one input voice sample a2 is input to the input voice buffer 1 and stored (step S11). The count value of 3 is incremented by 1 (step S1
2) Return. That is, the input audio samples a2 are input to and accumulated in the input audio buffer 1 one by one in synchronization with the sampling clock k2, and the accumulated number is counted by the audio sample counter 3 which counts the sampling clock k2.

The above-described voice sample number increasing process (step S5) is executed by the voice sample number converter 4 as a subroutine of the procedure as shown in FIG. In this increase processing, first, P- (M * N) voice samples, which are less than the predetermined number P, are taken out from the input voice buffer 1 and oversampled L times (step S21), (P -(M * N)) * L Divide into sample points. Then, for each of the subdivided sample points (L−M), a new voice sample is interpolated from the original voice sample that is L * N times adjacent and output to the input voice buffer 1 to be overwritten and stored (step S2
2) Furthermore, for each L subdivided sample points, a voice sample is output to the input voice buffer 1 P- (L * N) times and is overwritten and stored (step S23). That is, (P-M *
From the N) audio sample sequence, increase the number of audio samples by the above-mentioned oversampling and interpolation, and add a total of (L * N) + (P- (L * N)) = P audio samples to the input audio buffer. Restore to 1.

Here, the process for increasing the number of voice samples will be described in more detail as shown in FIG.
Although FIG. 6 shows a part of the audio sample sequence, A0 to A6 in the figure are audio samples at 8 kHz before oversampling, and points marked with X are L times (16 in this example).
X) Oversampled subdivided sample points. In addition, B0 to B6 in the figure are subdivided sample points (L
-M) positions for each, and new voice samples at these positions are obtained by interpolating from adjacent original voice samples (for example, A0 and A1 for the B1 point), and these new voice samples are L *. Output to the input voice buffer 1 N times,
After that, although not shown in the figure, the subsequent audio samples are divided into P- (L *
N) Output to the input voice buffer 1 times, and store a total of P voice samples in the input voice buffer 1 again.

Note that although L = 16 in this example, if P = 160, M = 1, and N = 4, then 156 voice samples before conversion are converted to (L-
It outputs to the input voice buffer 1 every (M * 15) (L * N = 16 * 4 = 64) times, and (P-L *) every (L = 16).
N = 160−16 * 4 = 96) times, output to the input speech buffer 1 finally 64 + 96 = 160 speech samples (8 kHz sampling) are re-stored in the input speech buffer 1 and the speech encoder 2 Is output to. That is, by outputting the voice samples obtained by interpolation from the voice sample sequence for every 15 subdivided sample points 64 times, in this section, every 15 subdivided sample points are dispersed to output 64 voice samples. However, the number of voice samples is increased by four.

The process of interpolating the voice sample at the new sample point from the original voice sample is performed by linear interpolation in this example. For example, the voice sample of the new sample point B2 is (the voice sample of A1 * 2/1
6) + (voice sample of A2 * 14/16). Note that other interpolation methods can of course be used in the present invention. It is conceivable to use linear interpolation as a simple interpolation method, but it is desirable to use interpolation by a sampling function in order to maintain the continuity of speech better. Further, in this example, as shown in FIG. 7A, the audio sample is output to the input audio buffer 1 64 times for each 15 subdivided sample points from the head position of the 16-times oversampled audio sample sequence. After that, the audio sample is output to the input audio buffer 1 96 times for every 16 subdivided sample points. In the present invention, for example, as shown in FIG. The order of the output and the output of
As shown in, the output of every 15 pieces or the output of every 16 pieces may be divided and dispersed. Note that in the method shown in FIG. 7C, the audio samples of new sample points are interpolated even in the interval of outputting every 16 pieces, but in the method shown in FIGS. 7A and 7B. , In the interval of outputting every 16 units, it coincides with the original sample point before oversampling, so that the original voice sample may be output without performing the interpolation process, which is preferable.

The above-mentioned voice sample number reduction processing (step S6) is executed by the voice sample number converter 4 as a subroutine of the procedure shown in FIG. In this reduction process, first, P which is larger than the predetermined number P is set.
+ (M * N) voice samples are taken from the input voice buffer 1, over-sampled L times (step S31), and divided into (P + (M * N)) * L subdivided sample points. . Then, for each of the subdivided sample points (L + M), a new audio sample is interpolated from the original audio sample that is L * N times adjacent, and is output to the input audio buffer 1 to be overwritten and stored (step S32). Further, for each L subdivided sample points, P- (L * N) times audio samples are output to the input audio buffer 1 and are overwritten and stored (step S33). That is, from the (P + M * N) speech sample sequence, the speech samples are reduced by the above oversampling and interpolation processing, and the total (L *
N) + (P− (L * N)) = P voice samples are stored again in the input voice buffer 1.

Here, the process for reducing the number of voice samples will be described in more detail as shown in FIG.
Although FIG. 9 shows a part of the audio sample sequence, A0 to A6 in the figure are audio samples at 8 kHz before oversampling, and the points marked with X are L times (16 in this example).
X) oversampled segmented audio sample points. In addition, B0 to B6 in the figure are positions for each of the subdivided sample points (L + M), new audio samples at these positions are obtained by interpolating from the adjacent original audio samples, and these new audio samples are obtained. Is input to the audio buffer 1, and subsequent audio samples are input P- (L * N) times for each L subdivided sample points, which is not shown in the figure. Output to audio buffer 1 and re-store a total of P audio samples in input audio buffer 1.

In this example, L = 16, but if P = 160, M = 1, and N = 4, then 164 voice samples before conversion are converted into (L +
It is output to the input voice buffer 1 every (M * 17) (L * N = 16 * 4 = 64) times, and every (L = 16) (P-L *).
N = 160−16 * 4 = 96) times, output to the input speech buffer 1 finally 64 + 96 = 160 speech samples (8 kHz sampling) are re-stored in the input speech buffer 1 and the speech encoder 2 Is output to. That is, by outputting the voice samples obtained by interpolation from the voice sample sequence for every 17 subdivided sample points 64 times, in this section, every 17 subdivided sample points are dispersed to output 64 voice samples. However, four voice samples are reduced.

Here, the above-mentioned interpolation processing is the same as the above-mentioned increase processing. Further, in this example, as shown in FIG. 10, after the audio sample is output to the input audio buffer 1 64 times for every 17 subdivided sample points from the head position of the 16-times oversampled audio sample sequence, The audio sample is output to the input audio buffer 1 96 times for every 16 sampled sample points. However, in the present invention, the reduction processing is performed in various modes as shown in FIGS. 7B and 7C, for example. be able to.

As described above, the process of increasing or reducing the number of audio samples is performed by the audio sample counter 3 while the audio encoder 2 outputs the audio encoded information bits e2 for one frame. The number of audio samples input to 1 is counted, and the number of audio samples input to the input audio buffer 1 (that is, the number of audio samples output to the audio encoder 2) is changed from an appropriate value due to the fluctuation of the bit clock d2. It is done when it is out of alignment. Therefore, even when the bit clock d2 fluctuates, the audio encoder 2 can be supplied with an appropriate number of audio samples adapted to its output rate, and the above-mentioned increase processing and reduction processing are distributed. Therefore, the continuity of the voice sample sequence can be maintained.

Next, the processing by the above speech decoding apparatus will be described with reference to FIGS. 11 to 14. In this audio decoding device, a digital data input interrupt process for inputting audio coded information bits h2 to the audio decoder 6 and converting the number of reproduced audio samples stored in the output audio buffer 5 is performed by the processing procedure shown in FIG. In addition, a voice sample output interrupt process for outputting the voice sample j2 from the output voice buffer 5 is performed by the processing procedure shown in FIG.

First, the digital data input interrupt processing shown in FIG. 11 is generated at each falling edge of the bit clock g2, and one bit of digital data (voice coded information bit h2) is input to the voice decoder 6 (step S41). ).
Then, the level of the frame synchronization signal f2 is confirmed (step S42), and if the level is not "L", the process returns (ends the interrupt routine) while the level is "L".
If so, it is a delimiter of frame processing, and therefore the following processing is continued in order to adjust the number of reproduced audio samples output from the output audio buffer 5 as necessary.

That is, the count value of the voice sample counter 7 is confirmed (step S43), and when the count value is a predetermined value P (160 samples in this example) that matches the input rate to the voice decoder 6, Since it is not necessary to adjust the number of reproduced voice samples, the count value of the voice sample counter 7 is cleared (step S47) and the process returns.
On the other hand, if the count value is not the predetermined value P, it does not match the current input rate to the audio decoder 6 due to the fluctuation of the bit clock g2, so the number of reproduced audio samples output from the output audio buffer 5 is adjusted. In order to do so, the converter 8 determines whether the count value is larger than the predetermined value P (step S44).

As a result, if the count value is larger than the predetermined value P (in this example, P + (M * N)), the number of audio samples to be increased, which will be described later, is applied to reproduce audio that is insufficient for audio reproduction. The sample is supplemented (step S45), and the count value is smaller than the predetermined value P (in this example, P-
(M * N)), the audio sample number reduction process described later is performed to reduce the reproduced audio samples remaining for audio reproduction (step S46), and the accumulated audio samples are stored in the output audio buffer 5 as 1 The number of reproduced audio samples of the frame is adjusted to a predetermined number P suitable for audio reproduction, and the P audio sample strings j2 are output from the output audio buffer 5 to an audio reproducer (not shown). Then, the count value of the audio sample counter 7 is cleared for the next frame processing (step S47), and the process returns.

The audio sample output interrupt processing shown in FIG. 12 is generated at each falling edge of the sampling clock l2, one reproduced audio sample j2 is output from the output audio buffer 5 (step S51), and the audio sample counter 7 is output.
The count value of is incremented by 1 (step S52), and the process returns. That is, the reproduced voice samples j2 are output from the output voice buffer 5 one by one in synchronization with the sampling clock l2, and the number of outputs is counted by the voice sample counter 7 that counts the sampling clock l2.

The above-described voice sample number increasing process (step S45) is performed by the voice sample number converter 8 as shown in FIG.
It is executed as a subroutine of the procedure as shown in. In this increase processing, first, the count value (P + (M *
N)) less than P reproduced voice samples are taken out from the output voice buffer 5, oversampled L times (step S61), and divided into P * L subdivided sample points. Then, for each of the subdivided sample points (L−M), a new voice sample is interpolated from the original voice sample that is L * N times adjacent to the output voice buffer 5
, And overwrite-store it (step S62). Further, for each L subdivided sample points, P + (M *
N)-(L * N) times output to the output voice buffer 5 and store it by overwriting (step S63).

That is, in the same manner as shown in FIG. 6 and FIG. 7, the reproduced sound samples are increased by the above-mentioned oversampling and interpolation from the P reproduced sound sample string, and the total (L * N) + ( P + (M * N)-(L * N))
= P + (M * N) reproduced voice samples are stored again in the output voice buffer 5. As a result, the output audio buffer 5 outputs a number of reproduced audio samples suitable for the audio reproduction process. The mode of increasing the reproduced audio samples is the same as the case of the above encoding.

The above-mentioned voice sample number reduction processing (step S46) is executed by the voice sample number converter 8 as a subroutine of the procedure shown in FIG. In this reduction processing, first, the count value (P
-(M * N)) more than P reproduced voice samples are taken out from the output voice buffer 5, oversampled L times (step S71), and divided into P * L subdivided sample points. Then, for each of the subdivided sample points (L + M), a new audio sample is interpolated from the original audio sample that is L * N times adjacent, and is output to the output audio buffer 5 to be overwritten and stored (step S72). , And P-the audio sample for each L subdivided sample points.
The data is output to (M * N)-(L * N) times output audio buffer 5 and is overwritten and stored (step S73).

That is, in the same manner as shown in FIG. 9 and FIG. 10, a total of (L * N) + (P -(M * N)-(L * N))
= P- (M * N) reproduced voice samples are stored again in the output voice buffer 5. As a result, the output audio buffer 5 outputs a number of reproduced audio samples suitable for the audio reproduction process. The mode of reducing the reproduced voice samples is the same as the case of the above encoding.

As described above, in the process of increasing or decreasing the number of voice samples, the voice sample counter 7 outputs the voice buffer 5 while the voice codec 6 inputs one frame of voice coded information bits h2. The number of reproduced audio samples output from the output audio buffer 5 is counted, and the number of reproduced audio samples input to the output audio buffer 5 (that is, the number of reproduced audio samples output to the audio reproducer) changes from an appropriate value due to the fluctuation of the bit clock g2. It is done when it is out of alignment.
Therefore, even when the bit clock g2 fluctuates, it is possible to supply an appropriate number of reproduced audio samples suitable for the reproduction process to the audio reproducer, and the above-mentioned increase processing and reduction processing are dispersed. Therefore, it is possible to reproduce high-quality sound while maintaining the continuity of the reproduced sound sample sequence.

In the above-described embodiment, the processing for increasing and reducing the number of audio samples is performed by reading from the audio buffer and overwriting again, but in the present invention, for example, it is stored in the input area of the audio buffer. After reading the converted audio sample and converting it, write it in the output area of the audio buffer and output it, or convert the number of samples with the converter when storing in the audio buffer or outputting from the audio buffer However, the aspect is not particularly limited. Further, in the present invention, there is no particular limitation on the mode of duplication or decimation in the process of increasing or reducing the number of audio samples, and the point is that the duplication or decimation may be performed by distributing the oversampled sample sequence without bias. .

[0056]

As described above, according to the present invention,
Since the number of audio samples supplied from the audio buffer to the audio encoder and the number of reproduced audio samples output from the audio buffer are distributed and adjusted without bias in the frame, for example, input from an external device. Even if there is a bit clock frequency error, the number of audio samples can be converted to an appropriate number for encoding and decoding playback, and the continuity between the audio samples can be maintained. It is possible to reproduce high quality sound.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a speech encoding apparatus according to an embodiment of the present invention.

FIG. 2 is a configuration diagram of a speech decoding apparatus according to an embodiment of the present invention.

FIG. 3 is a flowchart showing an example of a procedure of digital data output interrupt processing in voice encoding.

FIG. 4 is a flowchart showing an example of a procedure of a voice sample input interrupt process in voice encoding.

FIG. 5 is a flowchart showing an example of a procedure of a voice sample number increase process in voice encoding.

FIG. 6 is a conceptual diagram illustrating a voice sample number increase process.

FIG. 7 is a conceptual diagram illustrating a distribution mode in a voice sample number increase process.

FIG. 8 is a flowchart illustrating an example of a procedure of a voice sample number reduction process in voice encoding.

FIG. 9 is a conceptual diagram illustrating a voice sample number reduction process.

FIG. 10 is a conceptual diagram illustrating a distribution mode in a voice sample number reduction process.

FIG. 11 is a flowchart showing an example of a procedure of digital data input interrupt processing in audio decoding / playback.

FIG. 12 is a flowchart showing an example of the procedure of an audio sample output interrupt process in audio decoding and reproduction.

FIG. 13 is a flowchart showing an example of the procedure of a process for increasing the number of audio samples in audio decoding / playback.

FIG. 14 is a flowchart illustrating an example of a procedure of a voice sample number reduction process in voice decoding and reproduction.

[Fig. 15] Fig. 15 is a configuration diagram of a conventional speech encoding device and speech decoding device.

FIG. 16 is a timing chart showing an example of input / output timings in voice encoding and voice decoding.

[Explanation of symbols]

1 ... Input voice buffer, 2 ... Voice encoder,
3 ... voice sample counter, 4 ... voice sample number converter, 5 ... output voice buffer, 6 ... voice decoder, 7 ... voice sample counter, 8 ...
Voice sample number converter, a2, b2 ... voice sample, k2, 12 ... sampling clock, c2,
f2 ... Frame synchronization signal, d2, g2 ... Bit clock, e2, h2 ... Speech coded information bit,
i2, j2 ... Playback audio sample,

Claims (8)

(57) [Claims] 1. An audio sample is accumulated in an audio buffer for each predetermined frame in synchronization with a sampling clock, the accumulated audio samples are input to an audio encoder for each frame, and an output of the audio encoder is output. In the audio processing method that outputs the audio coded information bit string for each frame in synchronization with the bit clock, the audio sample for each frame accumulated in the audio buffer is oversampled and the bit clock is relative to the sampling clock. If the number of voice samples for each frame is input to the voice encoder by interpolating the voice sample at the new sample point from the original voice sample by lengthening the interval between sample points, Convert to the number of audio samples that matches the bit clock, and set the bit clock for the sampling clock. When the lock is relatively increased and changed, the interval between sample points is shortened and the audio sample at the new sample point is interpolated from the original audio sample, so that the audio for each frame input to the audio encoder is An audio processing method, characterized in that the number of samples is converted into the number of audio samples suitable for the bit clock. 2. An input voice sample is stored in a voice buffer for each frame (P pieces / frame), and the voice sample stored in the voice buffer is subjected to a voice encoding process by a voice encoder to obtain a bit clock. In a voice processing for outputting a voice coded information bit string (Q bits / frame) in synchronization with, the number of voice samples stored in the voice buffer while the voice coded information bits are output by Q bits is counted, When the count value is (P + (M * N)), the audio samples stored in the audio buffer are oversampled L times to obtain (P + (M * N)) * L subdivided samples. After setting the points, subdivided sample points (L +
Every (M) number of (L * N) times, a new voice sample is interpolated from the adjacent original voice sample and output to the voice encoder, and (P- (L *
N)) times output the audio sample to the audio encoder, so that a total of P audio samples are output to the audio encoder and the count value is (P- (M * N)). Oversamples the audio samples stored in the audio buffer by L times to obtain (P- (M * N)) * L subdivided sample points, and then subdivides the sample points (L-
Every (M) number of (L * N) times, a new voice sample is interpolated from the adjacent original voice sample and output to the voice encoder, and (P- (L *
N)) A speech processing method characterized in that a total of P speech samples are output to the speech encoder by outputting the speech samples to the speech encoder. 3. A voice coded information bit string for each frame is input to a voice decoder in synchronization with a bit clock, decoded playback voice samples are output, and the playback voice samples for each frame are stored in a voice buffer. Then, in the audio processing method that outputs the reproduced audio sample for each predetermined frame in synchronization with the sampling clock, the reproduced audio sample for each frame accumulated in the audio buffer is oversampled, and the bit clock is compared with the sampling clock. If there is a relative decrease, the interval between sampling points is shortened and the audio sample at the new sample point is interpolated from the original audio sample to determine the number of reproduced audio samples for each frame output from the audio buffer. Convert to the number of audio samples that matches the sampling clock, and When the bit clock relatively increases and changes with respect to the ringing clock, the interval between sampling points is lengthened and the audio sample at the new sample point is interpolated from the original audio sample and output from the audio buffer. An audio processing method, characterized in that the number of reproduced audio samples for each frame is converted into the number of audio samples suitable for the sampling clock. 4. Synchronized with a bit clock, a speech coded information bit string (Q bits / frame) is input to a speech decoder for each frame, speech decoding processing is performed for each frame, and reproduced speech signal samples (P / Frame) is stored in the audio buffer, and then in the audio processing of outputting the reproduced audio sample for each frame, the number of reproduced audio samples output from the audio buffer is counted while Q bits of audio encoded information bits are input. , If the count value is (P + (M * N)),
P output from the audio decoder and stored in the audio buffer
After oversampling the reproduced speech samples L times into (P * L) subdivided sample points, each subdivided sample point (L−M) is adjacent (L * N) times. A new audio sample is interpolated from the audio sample of
By outputting the audio sample from the audio buffer (P + (M * N)-(L * N)) times for each piece, a total of (P +
When (M * N)) reproduced voice samples are output and the count value is (P- (M * N)),
P output from the audio decoder and stored in the audio buffer
After oversampling each reproduced voice sample by L times to make (P * L) subdivided sample points, each subdivided sample point (L + M) (L * N) times the adjacent original voice A new audio sample is interpolated from the sample and output from the audio buffer, and the sample point L
By outputting the audio samples from the audio buffer for each (P- (M * N)-(L * N)) times, a total of (P-
A sound processing method comprising outputting (M * N)) reproduced sound samples. 5. A voice buffer for storing input voice samples, and a voice encoder for encoding the voice samples stored in the voice buffer, the voice encoder synchronizing with a bit clock. In a voice processing device that outputs a voice coded information bit string for each frame, a voice sample counter that counts the number of voice samples stored in a voice buffer while one frame of voice coded information bits is output from a voice encoder. And the count value by the audio sample counter is different from the predetermined number of audio samples for one frame to be stored in the audio buffer while one frame of audio encoded information bits is output from the audio encoder. An audio sample number converter for converting the number of audio samples stored in the audio buffer. The pull number converter oversamples the audio samples for each frame stored in the audio buffer, and when the count value is greater than the predetermined number of samples, the interval between sampling points is lengthened to increase the sampling point of the new sample points. By interpolating the voice samples from the original voice samples, the voice samples for each frame input to the voice encoder are converted into the predetermined number of samples, and sampling is performed when the count value is less than the predetermined number of samples. By shortening the point interval and interpolating the audio sample of the new sample point from the original audio sample,
A speech processing apparatus, characterized in that the speech samples for each frame input to the speech encoder are converted into the predetermined number of speech samples. 6. A voice decoder comprising: a voice decoder for voice-decoding a voice coded information bit string for each frame; and a voice buffer for storing and outputting reproduced voice signal samples that have undergone voice decoding for each frame. In a speech processing apparatus for synchronously inputting a speech coded information bit string to a speech decoder, the number of reproduced speech samples output from a speech buffer while one frame of speech coded information bits is inputted to the speech decoder. The audio sample counter for counting and the count value by the audio sample counter are the predetermined reproduced audio for one frame to be stored in the audio buffer while the audio encoded information bits for one frame are input to the audio decoder. An audio sample number converter that converts the number of audio samples stored in the audio buffer when the number of samples is different The audio sample number converter oversamples the reproduced audio samples for each frame stored in the audio buffer, and lengthens the sampling point interval when the count value is greater than the predetermined number of samples. By interpolating the audio sample at the new sample point from the original audio sample, the number of reproduced audio samples for each frame output from the audio buffer is converted to the predetermined number of samples, and the count value is the predetermined number of samples. If it is smaller, the interval between sampling points is shortened and the audio sample at the new sample point is interpolated from the original audio sample to determine the number of reproduced audio samples for each frame output from the audio buffer by the predetermined audio sample. A voice processing device characterized by converting into a number. 7. A voice communication device for encoding and transmitting a voice signal and decoding the received coded voice signal, wherein the voice processing device according to claim 5 is provided in a voice signal encoding processing section. A voice communication device comprising the voice processing device according to item 6 in a voice signal decoding processing unit. 8. A voice communication system in which a transmission side device encodes and transmits a voice signal, and a reception side device decodes a received encoded voice signal, wherein the voice processing device according to claim 5 is used. 7. A voice communication system comprising: the voice signal encoding processing unit according to claim 6; and the voice processing apparatus according to claim 6 included in a voice signal decoding processing unit of a reception side device.