JP3001584B2 - Audio signal transmission method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Purpose of the Invention] (Industrial application field) The present invention distinguishes a speech signal into a sound section with speech, a noise section with background noise, and a silent section, and The present invention relates to an audio signal encoding method that enables effective use of a line and reproduction of audio with excellent naturalness by encoding information of a noise section.

(Prior Art) A speech signal, especially a speech voice signal, is composed of a voiced section and a non-voice section with voice, and information as a conversational voice is included only in a voiced section. Therefore, when an audio signal is encoded and transmitted, a method for effectively using a line is conventionally adopted by transmitting only a code in a sound section.

FIGS. 10 and 11 show a conventional audio signal transmission / reception system based on such a system. FIG. 10 shows the configuration on the transmission side, and FIG. 11 shows the configuration on the reception side. On the transmitting side, the speech signal input by the sound detector 101 is distinguished into a sound section and a silent section, and only the sound section is encoded by the encoder 102 and transmitted.

On the other hand, on the receiving side, the received code is decoded by the decoder 111 into speech. The receiving side further includes a noise synthesizer 112 including a white noise generator. Switch 113 is a decoder
When the voice signal of the voiced section from the voice section 111 exists, it is output as it is, but when the voice signal of the voiced section does not exist, that is, in the voiceless section, the white noise from the noise synthesizer 112 is converted into the background noise of the voiceless section. Output as

However, in this method, the white noise from the noise synthesizer 112, which is output as background noise in a silent section on the receiving side, is different from the actual background noise that can be heard together with the voice in the voiced section, so that the voiced section and the silent Background noise greatly changes in the section, resulting in a very unnatural voice.

(Problems to be Solved by the Invention) As described above, according to the conventional technique, the noise information is not transmitted from the transmitting side, and white noise is output as background noise in a silent section on the receiving side. The background noise heard overlapping with the background noise in the silent section sounds very different. For this reason, the background noise changes between the sound section and the silent section, and the sound sounds unnatural.

The present invention solves such a problem, and reproduces reproduced sound with little change in background noise between a sound section and a silent section on the reception side without increasing the amount of transmission information from the transmission side. It is an object of the present invention to provide an audio signal encoding method for obtaining the same.

[Configuration of the Invention] (Means for Solving the Problems) In order to achieve the above object, the present invention determines a voiced section and a silent section of a voice signal, and performs voice information of a voiced section and noise information of a voiceless section. And transmitting the noise information after encoding the noise information at a higher compression ratio than the audio signal.

According to a second aspect of the present invention, in the sound signal transmitting method for determining a sound section and a silent section for a sound signal, and also transmitting sound information of the sound section and noise information of the silent section, the noise information is determined in advance. Only the noise information higher than the threshold value is encoded at a higher compression ratio than the voice signal and transmitted.

A third invention is a sound signal transmission method for determining a sound section and a sound section for a sound signal, and also transmitting sound information of the sound section and noise information of the sound section, wherein the noise information has a characteristic noise. It is characterized in that only information is encoded and transmitted at a higher compression ratio than the audio signal.

According to a fourth aspect of the present invention, there is provided an audio signal transmitting method for determining a sound section and a silent section with respect to an audio signal, and also transmitting voice information of the voice section and noise information of the silent section. A speech signal transmission method comprising: determining whether there is significant noise to be transmitted to the speech signal, encoding only the significant noise at a higher compression ratio than the speech signal, and transmitting the encoded speech signal.

According to a fifth aspect of the present invention, in the audio signal transmitting method for determining a voiced section and a silent section for a voice signal and also transmitting voice information of the voiced section and noise information of the voiceless section, the noise information of the voiceless section is predetermined. It is characterized in that it is determined whether or not there is noise exceeding a given threshold, and only significant noise is encoded at a higher compression ratio than the voice signal and transmitted.

According to a sixth aspect of the present invention, there is provided an audio signal transmitting method for determining a sound interval and a silent interval for an audio signal, and also transmitting voice information of a voice interval and noise information of a silent interval. Is determined to be significant noise in which noise can be heard, and only significant noise is encoded at a higher compression ratio than the voice signal and transmitted.

(Operation) Background noise in a silent section does not have an important meaning in conversation, but is necessary for natural conversation. Also, in a sound section, the shorter the transmission delay time is, the better to maintain the naturalness of conversation. However, since background noise only needs to be able to reproduce the characteristics of the noise, the amount of information can be compressed and transmitted from a sound section. Also, there is no strict delay time as in the sound section.

According to the present invention, by transmitting not only information on a sound section but also information on background noise in a silent section from the transmitting side, there is almost no change in background noise between a sound section and a silent section on the receiving side. Natural playback sound can be obtained.

In addition, the background noise can be encoded at a higher compression rate than the voiced frame by allowing the delay of the background noise on the transmission side, so that the amount of information caused by transmitting the background noise information hardly increases.

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of one embodiment of the present invention. In FIG. 1, reference numeral 10 denotes an audio signal input terminal through which a digital audio signal is input.

Buffers 132, 142, 150, 156 used below
Stores audio signal information in frame units, and it is necessary to store the order of the information input to the buffer at this time. The information of the audio signal is a feature parameter or a sample. In order to preserve the order relation accumulated in the buffer, the characteristic parameters are accumulated from the head of the buffer toward the tail, for example, in the order in which the characteristic parameters are input to the buffer. That is, the newest feature parameter (the feature parameter of the frame to be determined at present)
Is stored in the head of the buffer, and the last characteristic parameter is stored in the tail.

The voice detector 11 receives a digital voice signal as input, sets N samples of the digital voice signal as a frame, and discriminates, for each frame, a voiced section (voiced frame) in which voice exists. Note that the frame length N does not need to be constant.

The method is, for example, the linear prediction coefficient (LPC
Coefficient), autocorrelation coefficient, number of zero crossings, PARCOR coefficient, spectrum obtained by spectrum analysis, etc., are significantly different between voiced and non-voiced. can do.

FIG. 2 shows an example of a method for discriminating between sound and silence.

The power calculator 111 measures the average power P of one frame of the digital audio signal input from the audio signal input terminal 10 as follows. Here, X (i) is a sampled audio signal.

The threshold value comparator 112 compares P calculated by the power calculator 111 with a preset threshold value TP, and determines the presence or absence of sound as follows.

If P <TP, there is no sound. If P ≧ TP, sound is generated. The voice encoder 12 determines that the voice signal input through the voice signal input terminal 10 is voiced by the voice detector 11. When it is determined, the frame or the vicinity of the frame is encoded.

Here, coding is performed by a coding method such as APC-MLQ (adaptive predictive coding with maximum likelihood quantizer), ATC-VQ (adaptive transform coding with vector quantizer), and ADPCM. here,
It is not necessary to keep this code length constant.

At this time, basically, the audio signal is encoded in units of frames used in the sound detector 11 to be a code block. The audio signal of the frame unit used in the sound detector 11 may be encoded as an integral frame to form a code block. Further, the audio signal may be encoded in units of an integer fraction of the frame used in the sound detector 11, and may be used as a code block.

Here, information indicating that the block is a code block of a sound frame may be added to the head of the code string.

Further, identification information indicating the beginning and end of the code block may be added before and after the code block.

The silent encoder 13 encodes a frame determined by the sound detector 11 as not having sound.

FIG. 3 shows an example of the configuration. First, the encoder 131 encodes noise of a frame of the audio signal determined to be silent by the sound detector 11.

At the same time, the frame counter 130 counts the number of encoded frames, and issues an instruction to output a code to the output device 133 when the number of encoded frames is equal to or more than M frames. When the output device 133 outputs the code, the output device 133 issues a reset command to the frame counter 130. The frame counter 130 receiving the reset instruction starts counting the number of frames from “0”.

On the other hand, the code encoded by the encoder 131 is stored in the buffer 1
Stored in 32.

When the code enters the buffer 132 continuously for N frames, the buffer monitor 134 issues a command to output the code of the N frames in the buffer 132 to the output device. If a frame entering the buffer 132 is interrupted on the way, the buffer 132 is cleared.

The output device 133 outputs the code in the buffer 132 for N frames only when receiving an output command from both the frame counter 130 and the buffer monitor 134.

Here, if N >> N, there is almost no deterioration in efficiency due to transmission of background noise information.

The encoder 131 encodes the background noise. This encoder 1
31 may be the same as the speech encoder 12, or may be another to increase the compression ratio. However,
The use of a quantization table or the like suitable for background noise can improve the coding efficiency.

Thus, according to the present embodiment, when noise in the background of speech is perceived in a sound section, natural speech can be reproduced by transmitting the characteristics of the noise even in a silent section. Also, by separating the encoder for the signal in the silent section from that for speech,
Efficient encoding becomes possible. Moreover, since no significant noise detector is used, the hardware can be small. Further, unlike a speech signal, a noise signal does not need to be decoded accurately.

FIG. 4 is a block diagram showing a configuration of a speech signal encoding device according to a second embodiment of the present invention. This speech signal encoding device includes a sound detector 11, a significant noise detector 14, and a speech noise encoder 15.

FIG. 5 is a block diagram showing the configuration of the significant noise detector 14.

Here, it is checked whether or not significant noise as perceived on the receiving side is present in the frame, from among the frames determined to be non-sound by the sound detector 11.

Here, the determination as to whether the noise is significant is made based on the following characteristics of the audio signal.

The frame having the minimum average power can be determined to be silent. When the minimum value of the average power is larger than a certain value, noise in a silent section is heard on the receiving side. Such a noisy frame is determined to be transmitted and is determined to be a significant noise frame.

First, the in-frame power measuring device 141 calculates the average power in a frame of the audio signal, which is determined to be silent by the sound detector 11, in the frame.

At the same time, the frame counter 140 counts the number of frames input from the terminal, and outputs “1” to the determination means 145 when the number becomes M frames or more. When it is less than M, "0" is output. When the determination unit 145 outputs information indicating that the noise is significant, the determination unit 145
Issues a reset instruction to the frame counter. The frame counter 140 that has received the reset instruction starts counting the number of frames from 0.

On the other hand, the value of the average power in the frame measured by the intra-frame power measuring device 141 is stored in the buffer 142.

When the output of the intra-frame power measuring device 141 enters the buffer 142 for N consecutive frames, the buffer monitor 144
From the IIth in the buffer 142 to the in-buffer power meter 143
Calculate the average of the power of the JJ-th L frame (L = JJ-I
I), and issue a command to output the value to the determiner 145. If the frame entering buffer 142 is interrupted in the middle, buffer 1
Clear 42.

The determiner 145 outputs information indicating significant noise only when the output of the frame counter 140 is “1” and the output value of the in-buffer power measuring device 143 exceeds a certain threshold T. Output.

However, detection of significant noise may be performed without using the information of the sound detector 11.

The speech noise encoder 15 may be the same as the speech encoder 12. In that case, if it is either between sound frames or a significant noise frame, the frame is encoded and transmitted.

Further, even though the configuration of the encoder itself is the same, encoding parameters such as an encoding speed and a quantization table may be changed depending on speech and noise.

 FIG. 6 shows a configuration example of the speech noise encoder 15.

The audio signal from the audio input terminal 10 is stored in the buffer 150 as it is.

On the other hand, if the current frame is determined to be silent by the speech detector 11 and is determined to be significant noise by the significant noise detector 14, the coded frame determiner 151 stores the current frame in the buffer 150. It instructs the encoder 152 to encode the I-th to J-th frames from the current frame among the existing frames. The quantization table at that time is supplied to the noise quantization table 155 by the switch 153.
Is switched to

When the voiced detector 11 determines that the current frame is voiced, the coded frame determiner 151 determines, from among the frames stored in the buffer 150, the I-th to J-th frames from the current frame. To encode the frame of
An instruction is issued to the encoder 152. At this time, I = 0 and J = 0
It is. That is, the encoder 152 selects one of the 150 in the buffer.
The newest frame signal is encoded and output. The quantization table at that time is switched to the audio quantization table 154 by the switch 153.

The encoder 152 encodes the I-th to J-th frames from the current frame in the buffer 150. At this time, encoding may be performed in frame units, or the I-th to J-th frames may be encoded collectively. The encoding can be performed in the same manner as the speech encoder 12.

According to the present embodiment, a frame of significant noise can be accurately detected by the significant noise detector. Moreover, since only one encoder is required, the hardware is small.

 FIG. 7 shows another example of the configuration of the speech noise encoder 15.

This is basically the same in operation as the speech noise encoder of FIG. The difference is the buffer. Since the buffer 150 in FIG. 6 needs to store audio samples,
The size of the buffer increases. But the seventh
The buffer 156 shown in the figure only needs to store the code of the audio coded by the encoder 152, so that the size of the buffer 156 is a fraction of that of the buffer 150, so that the hardware scale is reduced. However, when significant noise is encoded, it must be encoded one frame at a time, which has the disadvantage that the compression rate is lower than that of the method of FIG.

FIG. 8 shows a third embodiment of the present invention, which comprises a sound detector 11, a speech encoder 12, a significant noise detector 14, and a significant noise encoder 16.

FIG. 9 is a block diagram showing the configuration of the significant noise encoder 16.

The buffer 160 accumulates an audio signal input from the audio input terminal 10 in frame units.

The coded frame determiner 161 determines that the frames from the II-th frame to the JJ-th frame among the frames stored in the buffer are to be coded.

The encoder 162 encodes the audio signal in the buffer 160. At this time, encoding may be performed in frame units,
Frames (K <N) may be encoded together.

On the receiving side, as a method for identifying whether a received code block is for speech or significant noise, when speech is encoded, "1" is added to the beginning of the code.
By coding “0” at the beginning of the code when significant noise is encoded, the receiving side can identify whether the code block is speech or significant noise.

According to the present embodiment, a frame of significant noise can be accurately detected by the significant noise detector. In addition, by separating the noise frame encoder from that for speech, it is possible to encode the noise frame with high efficiency.

[Effects of the Invention] As described above, according to the present invention, when noise in the background of speech is perceived in a sound section, a natural sound can be reproduced by transmitting the characteristics of the noise in a silent section.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the configuration of a speech signal encoding apparatus according to a first embodiment of the present invention, FIG. 2 is a block diagram showing the configuration of a sound detector 11, and FIG. 3 is a silence encoder. 13 is a block diagram showing the configuration of the second embodiment of the present invention, FIG. 5 is a block diagram showing the configuration of the significant noise detector 14, and FIGS. FIG. 8 is a block diagram showing a configuration of the noise encoder 15, FIG. 8 is a block diagram showing a third embodiment of the present invention, FIG. 9 is a block diagram showing a configuration of the significant noise encoder 16, FIG. FIG. 11 is a block diagram showing a configuration of a conventional audio signal transmitting / receiving system. 11 Voice detector 12 Voice encoder 13 Silence encoder 14 Significant noise detector 15 Voice noise encoder 16 Significant noise encoder 18 Transmit code output Terminal

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int. Cl. ⁷ , DB name) H04J 3/00 H04L 12/56 H04B 14/00

Claims (6)

(57) [Claims] 1. A voice signal transmitting method for determining a voiced section and a voiceless section of a voice signal and transmitting voice information of a voiced section and noise information of a voiceless section, wherein the noise information is compressed at a higher level than the voice signal. An audio signal transmission method characterized in that the audio signal is encoded at a rate and transmitted. 2. A voice signal transmitting method for determining a voiced section and a silent section of a voice signal, and also transmitting voice information of a voiced section and noise information of a voiceless section, wherein the noise information is determined in advance by a predetermined threshold. An audio signal transmission method, characterized in that only noise information higher than a value is encoded at a higher compression ratio than the audio signal and transmitted. 3. A voice signal transmitting method for determining a voiced section and a voiceless section of a voice signal and transmitting voiced information of a voiced section and noise information of a voiceless section. An audio signal transmission method characterized in that it is encoded at a higher compression ratio and transmitted. 4. A voice signal transmitting method for determining a voiced section and a voiceless section of a voice signal and transmitting voice information of a voiced section and noise information of a voiceless section, wherein the noise information of the voiceless section is transmitted to a receiving side. An audio signal transmission method, comprising determining whether there is significant noise to be transmitted, encoding the significant noise only at a higher compression ratio than the audio signal, and transmitting the encoded signal. 5. A voice signal transmitting method for determining a voiced section and a voiceless section of a voice signal and transmitting voice information of a voiced section and noise information of a voiceless section, wherein the noise information of the voiceless section is predetermined. An audio signal transmitting method, wherein it is determined whether or not there is noise exceeding a threshold value, and only significant noise is encoded at a higher compression ratio than the audio signal and transmitted. 6. A voice signal transmitting method for determining a voiced section and a voiceless section of a voice signal and also transmitting voice information of a voiced section and noise information of a voiceless section. An audio signal transmission method, comprising: determining whether significant noise is audible, encoding only the significant noise at a higher compression ratio than the audio signal, and transmitting the encoded audio signal.