JPH0236628A - Transmission system and transmission/reception system for voice signal - Google Patents

Abstract

PURPOSE:To attain the effective utilization of a line and the reproduction of voice with excellent naturality by sending a voice signal while taking the presence of a voice period and a silence period with background noise into account. CONSTITUTION:A voice/silence decision means 2 decides whether a signal is in the voice part or the silence part for each frame, the voice frame is outputted to a voice coding means 3 and the noise frame decided to the silence part is outputted to a buffer means 4 respectively. The voice coding means 3 adds identification information to the voice frame, the buffer means upon the storage of a prescribed number T of noise frame, transfers them to a noise coding means 5, where k-set of code blocks are generated and the identification information is added. The voice and noise coding blocks obtained by the voice coding means 3 and the noise coding means 5 are sent. Thus, not only the information at the voice frame but also the information of background noise in the silence frame are sent from the sender side. Thus, natural reproduced voice having almost no change in the background noise depending on the voice and silence periods is obtained at the receiver side.

Description

[Detailed Description of the Invention] [Objective of the Invention] (Industrial Application Field) The present invention eliminates the problem of line congestion by transmitting audio signals taking into account the presence of sound sections and silent sections with background noise. The present invention relates to an audio signal transmission method and a transmission/reception method that enable reproduction of audio with excellent efficiency and naturalness.

(Prior Art) An audio signal, particularly a conversational audio signal, consists of a sound section and a silent section, and information about the conversational sound is included only in the sound section. Therefore, when a voice signal is encoded and transmitted, a conventional method has been adopted in which only the codes of the voiced sections are transmitted, thereby making effective use of the line.

FIGS. 6 and 7 show conventional audio signal transmitting and receiving systems based on such a system, with FIG. 6 showing the configuration of the transmitting side and FIG. 7 showing the configuration of the receiving side, respectively. On the transmitting side, a speech detector 61 distinguishes the input audio signal into a speech section and a silent section, and an encoder 62 encodes and transmits only the speech section.

On the other hand, on the receiving side, the received code is decoded into speech by a decoder 71 (;). The receiving side is further equipped with a noise synthesizer 72 consisting of a white noise generator.

The switch 73 outputs the voice signal from the decoder 71 as it is when there is a voice signal in a voice zone, but when there is no voice signal in a voice zone, that is, in a silent zone, it outputs the white noise from the noise synthesizer 72. Output as background noise during silent periods.

However, in this method, the white noise from the noise generator 72 that is output as background noise during the silent period on the receiving side is different from the actual background noise that is heard together with the audio signal during the sound period. The background noise changes greatly between silent sections, resulting in extremely unnatural sound.

(Problem to be Solved by the Invention) In this way, in the conventional technology, the transmitting side does not transmit noise information, and the receiving side outputs white noise as background noise in the silent section. There was a problem in that the background noise between sections sounded very different.

The present invention solves these problems, and allows the receiving side to obtain reproduced audio with little change in background noise between the sound section and the silent section without significantly increasing the amount of information transmitted from the transmitting side. The purpose of this invention is to provide a method for transmitting and receiving audio signals.

[Structure of the Invention] (Means for Solving the Problems) The audio signal transmission method according to the present invention divides the audio signal into frames of a predetermined length in the voice/silence determining means, and determines whether there is a voice portion or a silent portion for each frame. A sound frame determined to be a sound part is outputted to a sound encoding means, and a noise frame determined to be a silent part is outputted to a buffer means. The voice encoding means encodes the voice frame to generate code blocks, and adds identification information to each code block indicating that it is a famous code block. When the buffer means accumulates a predetermined number T of noise frames, it transfers the T 21-tone frames to the noise encoding means. The noise encoding means encodes a predetermined frame selected based on a predetermined criterion among the T noise frames to generate T code blocks, and indicates to each code block that it is a noise code block. Add identification information. Then, the speech and noise code blocks obtained by the speech encoding means and the noise encoding means are transmitted.

Furthermore, in the audio signal transmission/reception system according to the present invention, code blocks of voice and noise are transmitted from the transmitting side as described above. On the receiving side that receives this, the code block separation means first determines whether each code block is a voice code block or a noise code block based on the identification information added to the code block, and then is output to the voice decoding means, and the code block of 2i tones is output to the noise decoding means.

Then, the natural sound signal and the noise signal obtained by the sound decoding means and the noise decoding means are concatenated and output.

(Function) In the present invention, the transmitting side not only transmits information on sound intervals but also information on background noise in silent intervals.
On the receiving side, natural reproduced audio with almost no change in background noise between the sound section and the silent section can be obtained.

Furthermore, on the transmitting side, sound frames are encoded frame by frame and transmitted, but noise frames are encoded all at once when T frames have been accumulated in the buffer means, and T frames are encoded at the same time as the number of frames is less than the number of frames. Sent as a code block. That is, noise frames are encoded at a higher compression rate than speech frames.

In order to maintain the naturalness of the conversation and to proceed with smooth conversation, it is better for the transmission delay time of the information in the voiced section to be as short as possible, but according to the present invention, since the voiced frame is encoded frame by frame, Transmission delay time is extremely small.

On the other hand, background noise during silent periods does not have an important meaning in transmitting information about the conversation content, but is necessary for natural conversation. However, since it is only necessary that the characteristics of background noise can be reproduced on the receiving side, it is possible to compress the amount of information from the sound interval and transmit it as in the present invention.

In addition, in the present invention, noise frames are transmitted to the receiving side with a transmission delay of T frames, but since most background noises are stationary, even if there is some transmission delay, it is natural. It will not damage.

(Example) Hereinafter, the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a transmitting side according to an embodiment of the present invention. In the figure, an audio signal to be transmitted is input to a voice/silence determiner 2 via an input terminal 1. The sound/silence determiner 2 divides the input audio signal into frames of a predetermined length, determines whether each frame is a sound part or a silent part, and separates the sound frame determined to be a sound part and the silent part. and the determined noise frame are separated and output.

The sound frames output from the sound/silence determiner 2 are 6
- input to g encoder 3; The four-tone encoder 3 converts the input voice frame into, for example, APC-MLQ (adaptive predictive coding with maximum quantization).

It is encoded using a coding method such as ATC-VQ (adaptive transform coding with vector quantization), and a code string of a certain length is used as a code block, and the beginning of the code block is marked as a voice code block. The identification information shown is added and output. Furthermore, when transmitting this code block via the transmitting circuit 6, information indicating the start and end of each No. 71 block may also be added. Here, there is no need for a 1:1 correspondence between a voice frame and a code block; for example, several sound frames may correspond to one code block, or one four-person frame may correspond to one code block. may correspond to multiple code blocks. Further, the length of the code block does not need to be constant and may be a variable length block.

On the other hand, the noise frame output from the voice/silence determiner 2 is input to the buffer 4. The buffer 4 holds a predetermined number of input noise frames. A noise encoder 5 is connected to the buffer 4. noise encoder 5
When T noise frames are accumulated in the buffer 4, it issues a transfer command to the buffer 4, collectively encodes the T noise frames transferred, and outputs k code blocks. Note that T does not necessarily have to be constant.

The sound and silence code blocks obtained from the sound coder 3 and the soundless coder 5 are sent from the transmitter 6 to the transmission path via the output terminal 7 and are transmitted.

FIG. 2 is a block diagram showing a specific example of the configuration of the noise encoder 5. As shown in FIG. The noise frame from buffer 4 in FIG.
The intra-frame feature detector 12 is inputted via a terminal 11 . This feature detector 12 analyzes and detects features within the noise frame and outputs a trace amount within the frame. This intra-frame time trace amount is transferred to the buffer 13 and accumulated. The inter-frame feature detector 14 detects inter-frame feature amounts from the intra-frame time traces accumulated in the buffer 13. Based on the intra-frame time trace amount accumulated in the buffer 13 and the frame opening time trace amount detected by the inter-frame feature detector 14, the determiner 15 selects j of the T noise frames.
Decide to encode frames. The determination result of the determiner 15 is applied via the buffer 13 to the encoder 16 to which the output of the intra-frame feature detector 12 is input. According to the determination result of the determiner 15, the encoder 16
The intra-frame time traces of j noise frames are encoded and output to the transmitter 6 in FIG. 1 via the terminal 17.

More specifically, each part of the noise encoder 5 shown in FIG. .

One feature such as a linear pre-ul coefficient is calculated. i-th feature in m-th noise frame] is defined as CAI(m)(
m-1,2,-, T%i-1,2,-■). The intra-frame time minute amount CAi(m) thus calculated is held by the buffer 13 from m-1 to T.

In the interframe feature detector 14, when the frame opening time mf;kc Ai(i) (m -1~T) is accumulated in the buffer 13, the correlation CI i(m
, n) is detected. For example, if CAi(a+) is the Fourier transform in the m-th noise frame, its first component C
A if) is the power of the m-th frame. As the interframe correlation, for example, CA i(m)(m-1~T)
and CA I (11) correlation CI i (m, n) (m-
1 to T s n-1 to T), etc. can be used.

The determiner 15 uses the intra-frame feature CA 1(i) stored in the buffer 13 and the inter-frame feature detector 14.
Inter-frame features (inter-frame correlation) from CI l
From (i, n), the number j of frames to be encoded among the T noise frames is determined. For example, the determination threshold T h
l, T h2, if CA 1(1)≧Tbl, then j = a+In [n l CI l(1, n) l
>Th2] ・=(1)CA 1(1)<T
If hl, determine j as follows: j-0...(2). However, formula (1) represents the minimum n that satisfies the conditions in [ ].

The encoder 16 encodes the features of the j noise frames from the beginning in the batch 13 as j code blocks based on j calculated by the determiner 15, and adds noise at the beginning of each code block. The code block is output with identification information indicating that it is a code block.

Furthermore, when transmitting this code block via the transmission circuit 6, information indicating the start and end of each code block is also added. In this way, a noise frame in a silent section is encoded and transmitted only when the significance of the noise is large, in other words, when the noise is at a level equal to or higher than that recognized as background noise.

FIG. 3 is a block diagram showing the configuration of the receiving side.

In the same figure, a signal transmitted from the transmitting side shown in FIG.

The signal received by the receiving unit 21 is sent to the code block separator 22
is input. The code block separator 22 distinguishes between code blocks and noise code blocks based on the identification information added to the manually input code blocks, and sends the voice code blocks to the voice decoder 23, and sends the voice code blocks to the voice decoder 23. are transferred to the buffer 24, respectively. The sound decoder 23 decodes the sound code block in a process reverse to that of the sound encoder 3 in FIG.

The buffer 24 temporarily stores input silence code blocks and transfers the stored noise code blocks to another buffer 25 according to instructions from the buffer monitor 26 . The buffer 25 stores noise code blocks to be decoded by the noise decoder 27, and transfers them to the noise decoder 27 according to instructions from the bus monitor 26.

The buffer monitor 26 stores k in the buffer 24.

If the next noise code block does not arrive even after waiting for a certain period of time Tb after the th noise code block has entered, the noise code blocks in the buffer 24 are transferred to the buffer 25. The number of noise code blocks stored in the buffer 25 at this time is k''. After the i-th noise code block in the buffer 25 is decoded, the i-th noise code block in the buffer 24 is They are transferred to the buffer 25 in the order of 1 to 2. Here, if k'>k'', and even after the last noise block in the buffer 25 has been decoded, there are still noise code blocks remaining in the buffer 24. If so, the remaining noise code blocks in the buffer 24 are immediately transferred to the buffer 25. Conversely, if k'<k'', even after all the noise code blocks in the buffer 24 have been transferred to the buffer 25, the previous noise code block remains in the buffer 25, so it is deleted. After that, Let k ”−k ′.

The noise decoder 27 decodes the noise code blocks accumulated in the buffer 25 from the beginning in a process reverse to that of the encoder 16 in FIG.
When the kth noise code block is decoded, decoding is performed again starting from the first noise code block.

The voice signal and the noise signal, which are the decoding results obtained by the voice decoder 23 and the noise decoder 27, respectively, are sent to the switch 2.
8 is input. If the switcher 28 receives a voice signal from the voice decoder 23, it is output as is, and if the decoding result from the voice decoder 23 is not input, it is output from the noise decoder 27. outputs a noise signal. That is, the switch 28 outputs the output of the voice decoder 23 and the noise decoder 2.
7 is connected and output. The output of this switch 28 is input to a smoothing filter 29 . The smoothing filter 29 connects the output of the 1 (°° code block) and the output of the 1st code block from the encoder, and connects the output of the q-tone decoder 'a23 and the output of the noise decoder 27. Smooth the area using the smoothing operation.

The outputs of these two smoothing filters are output from output terminal 30.
It is output as playback audio.

Next, another example of the configuration of the noise encoder 5 in FIG. 1 will be described with reference to FIGS. 4 and 5.

In the example shown in FIG. 2, the output of the intraframe feature detector 12 is input to the encoder 16 via the buffer 13, but in the example shown in FIG. The signal is directly input to the encoder 16 from the terminal 11.

In this case, the encoder 16 encodes j frames from the beginning of the human-generated noise frame using a method such as DPCM or orthogonal transform encoding, and divides the initial code into code blocks. Then, as described above, identification information indicating that the code block is a noise code block is added to the beginning of each code block and output.

In addition, in FIG. 2, the output of the intra-frame feature detector 12 is manually inputted to the inter-frame feature detector 14 via the bumper 13, but in FIG. Detection of intra-frame features and inter-frame features are performed in 1k steps.

In addition, the present invention can be implemented with various modifications without departing from the scope of the invention.

[Effects of the Invention] As described above, according to the present invention, encoded data related to background noise is transmitted even during a silent section, and the receiving side uses it to reproduce the background noise and output it during the silent section. Since there is almost no change in the stationary background noise between the sound section and the silent section, reproduced speech with excellent naturalness can be obtained.

Additionally, background noise information differs from the information of voiced sections in that it does not require very accurate reproduction, and even if there is a time delay, it does not cause any unnaturalness because it is mostly composed of stationary components. By focusing on T noise frames and transmitting them as code blocks smaller than H sound frames, the amount of transmitted information does not increase and the decrease in transmission efficiency is kept to a minimum. can.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of the transmitting side in one embodiment of the present invention, FIG. 2 is a block diagram showing a specific configuration example of the noise encoder in FIG. 1, and FIG. FIGS. 4 and 5 are block diagrams showing other configuration examples of the noise encoder according to the present invention. FIGS. 6 and 7 are diagrams showing the conventional transmission and reception of audio signals. FIG. 2 is a block diagram for explaining the method. DESCRIPTION OF SYMBOLS 1... Audio signal input terminal, 2... Speech/no-speech determiner, 3... Speech encoder, 4... Buffer, 5... Noise encoder, 6... Transmitter, 7... Transmission output terminal, 2
0... Reception input terminal, 21... Receiving section, 22...
Code block separator, 23...Speech decoder, 27...
- Noise decoder, 28... Switcher, 29... Smoothing filter, 30... Playback audio output terminal. Applicant's Representative Patent Attorney Takehiko Suzue Figure 1 Figure 5 Figure 6 Figure 2 Figure 7

Claims (2)

[Claims] (1) Divide the audio signal into frames of a predetermined length, determine whether each frame is a voiced portion or a silent portion, and separate and output the voiced frame determined to be a voiced portion and the noise frame determined to be a silent portion. A speech/non-speech determination means; a speech frame obtained by the speech/non-speech determination means is encoded to generate a code block, and each code block is provided with identification information indicating that it is a speech code block. a voice encoding means for adding; a buffer means for holding the noise frames obtained by the voice/non-speech determining means; and when a predetermined number T of noise frames are accumulated in the buffer means, the T noise frames are noise encoding means for generating k code blocks for a frame and adding identification information to each code block indicating that it is a noise code block; and means for transmitting the obtained voiced and noise code blocks. (2) Divide the audio signal into frames of a predetermined length, determine whether each frame is a voiced portion or a silent portion, and separate and output the voiced frame determined to be a voiced portion and the noise frame determined to be a silent portion. A speech/non-speech determination means; a speech frame obtained by the speech/non-speech determination means is encoded to generate a code block, and each code block is provided with identification information indicating that it is a speech code block. a voice encoding means for adding; a buffer means for holding the noise frames obtained by the voice/non-speech determining means; and when a predetermined number T of noise frames are accumulated in the buffer means, the T noise frames are noise encoding means for generating k code blocks for a frame and adding identification information to each code block indicating that it is a noise code block; means for transmitting the obtained voiced and noise code blocks; and a method for determining whether each code block transmitted by the means is a voiced code block or a noise code block based on identification information added to the code block. code block separation means for determining and separating and outputting voiced and silent code blocks; voice decoding means for decoding the voiced code blocks obtained by this means; and code block separation means for decoding the voiced code blocks obtained by this means; a noise decoding means for decoding a code block of noise; and a means for concatenating and outputting a voice signal and a noise signal obtained by the voice decoding means and the noise decoding means. transmission and reception method.