CN104022967A - Voice decoding apparatus - Google Patents

Abstract

An audio decoding device capable of reducing deterioration in call quality even when silent compression is applied. It includes: a jitter absorbing buffer for temporarily accumulating received packets and outputting the packets at a predetermined output timing; and a background noise generating unit for generating background noise data contained in packets output from the jitter absorbing buffer. noise audio data; an audio decoding unit that decodes audio coded data included in the packet output from the jitter absorbing buffer to generate voice audio data; a speech rate conversion unit that decodes the audio data decoded by the audio decoding unit Speech rate conversion that converts the reproduction speed of the above-mentioned audio data; and a control unit that controls the time length of the background noise generated by the background noise generation unit based on the accumulation status of packets in the above-mentioned jitter absorption buffer, and controls The reproduction speed converted by the speech rate conversion unit.

Description

audio decoding device

technical field

The present invention relates to an audio decoding device for decoding encoded audio used in Internet telephones and the like.

Background technique

Voice calls such as Internet telephony are conducted by encoding voice to form packets, and sending and receiving packets via the network. In packet communication, the time interval at which packets are received is often not fixed, and the time interval at which packets are received varies (jitters) in many cases. As a technique for absorbing such jitter and continuously outputting decoded audio obtained by decoding an audio code included in a received packet, there is a technique described in Patent Document 1, for example.

In the technique described in Patent Document 1, the reproduction speed is increased or decreased in accordance with the accumulation amount of reception packets in the jitter absorption buffer temporarily storing reception packets, thereby keeping the accumulation amount of reception packets in the jitter absorption buffer at Just the right amount, and output decoded audio continuously. As a result, the degradation of audio quality is reduced compared to the case where the accumulated amount of received packets is kept at an appropriate amount by discarding and duplicating received packets in the jitter absorbing buffer.

[Patent Document 1] Japanese Patent No. 3796240

However, in the conventional audio decoding device, the control is performed on the premise that the voice is encoded at fixed time intervals, the packets that are packaged and transmitted are stored in the jitter absorbing buffer corresponding to the packet number of the packet. at the location. Therefore, in a system that applies silent compression that does not necessarily send packets at constant time intervals, such as when the packet sending interval becomes longer in a silent period, there is a problem that appropriate processing cannot be performed, resulting in degradation of call quality.

Contents of the invention

The present invention was made in order to solve the above-mentioned problems, and an object of the present invention is to obtain an audio decoding device capable of reducing deterioration in speech quality even when silent compression is applied.

The audio decoding device of the present invention includes: a jitter absorbing buffer temporarily storing received packets and outputting the packets at a predetermined output timing; and a background noise generating unit based on background noise data for generating audio data for background noise; an audio decoding unit for decoding audio coded data included in the packet output from the jitter absorbing buffer to generate audio data for speech; a speech rate conversion unit for converting Speech rate conversion that converts the playback speed of the audio data decoded by the audio decoding unit; and a control unit that controls the background noise generated by the background noise generation unit based on the accumulation status of packets in the jitter absorbing buffer. and control the above-mentioned reproduction speed converted by the above-mentioned speech rate conversion section.

According to the present invention, by including: a jitter absorbing buffer temporarily storing received packets and outputting the packets at a predetermined output timing; The noise data generates audio data of background noise; the audio decoding unit decodes the audio coded data included in the packet output from the jitter absorbing buffer to generate audio data of speech; a speech rate conversion for converting the playback speed of the audio data decoded by the decoding unit; and a control unit for controlling the timing of the background noise generated by the background noise generation unit based on the accumulation status of packets in the jitter absorbing buffer. Length, and control the above-mentioned reproduction speed converted by the above-mentioned speed-of-speech conversion unit, thereby preventing deterioration of call quality even when silent compression is applied.

Description of drawings

FIG. 1 is a functional block diagram of an audio decoding device according to Embodiment 1 of the present invention.

FIG. 2 is an explanatory diagram showing the relationship between the time stamp of a packet and accumulation in a jitter absorbing buffer.

FIG. 3 is a functional block diagram of an audio decoding device according to Embodiment 2 of the present invention.

FIG. 4 is a functional block diagram of an audio decoding device according to Embodiment 3 of the present invention.

5 is a functional block diagram of an audio decoding device according to Embodiment 4 of the present invention.

FIG. 6 is a functional block diagram of an audio decoding device according to Embodiment 5 of the present invention.

FIG. 7 is an explanatory diagram showing the relationship between the time stamp of a packet and accumulation in a jitter absorbing buffer.

Label description

1: Jitter absorbing buffer; 2: Background noise generating unit; 3: Audio decoding unit; 4: Speech rate converting unit; 5: Output buffer; 6: Output buffer monitoring unit; 7: Control unit; 71: Buffer Surplus monitoring part; 72: Control signal output part; 73: Arrival speed monitoring part; 8: High-precision silent compression part; 9: Audio detection part; 10: Audio coding part; 11: Silent compression control part; 12: Background noise Data detection/insertion unit; 20: audio decoding device; 21: audio encoding device.

Detailed ways

Embodiments of the present invention will be described below. In addition, the following embodiment is an example of this invention, and this invention is not limited to the following embodiment.

Implementation mode 1.

FIG. 1 is a functional block diagram showing an audio decoding device according to an embodiment of the present invention.

In FIG. 1 , a jitter absorbing buffer 1 temporarily stores received packets, and outputs the packets at a predetermined output timing. The background noise generating unit 2 generates audio data of background noise based on the background noise data contained in the packets output from the jitter absorbing buffer 1 . The audio decoding unit 3 decodes the encoded audio data included in the packet output from the jitter absorbing buffer 1 to generate audio data of speech. The speech rate conversion unit 4 performs speech rate conversion for converting the playback speed of the audio data decoded by the audio decoding unit 3 . The output buffer 5 temporarily stores the audio data of the background noise generated by the background noise generating unit 2 and the audio data of speech generated by the audio decoding unit 3 . The output buffer monitoring unit 6 monitors the storage amount of the audio data stored in the output buffer 5, and instructs the jitter absorbing buffer 1 to output timing of the temporarily stored packets based on the storage amount. The control unit 7 controls the duration of the background noise generated by the background noise generation unit 2 and controls the playback speed converted by the speech rate conversion unit 4 according to the accumulation status of packets in the jitter absorption buffer 1 .

In the present embodiment, the control unit 7 has a buffer remaining amount monitoring unit 71 and a control signal output unit 72 . The buffer remaining amount monitoring unit 71 monitors the remaining amount of the jitter absorbing buffer 1 as the state of accumulation of packets in the jitter absorbing buffer 1 . The control signal output unit 72 outputs a time length control signal for controlling the time length of the background noise generated by the background noise generating unit 2 and a time length control signal for controlling the time length of the background noise generated by the background noise generating unit 2 based on the remaining amount of the jitter absorption buffer monitored by the buffer remaining amount monitoring unit 71. The reproduction speed control signal for controlling the reproduction speed converted by the speech rate conversion unit 4 .

Next, the operation will be described.

In addition, in the present embodiment, the operation in the case of a voice call between the user and the user's counterpart is described, but the present invention is not limited thereto.

First, when the user's counterpart speaks, the voice is encoded at the counterparty's side to form packets, and is received at the user's side through the network. In this way, when the user side receives a packet transmitted from the communication partner side, the jitter absorbing buffer 1 temporarily stores the received packet. The jitter absorbing buffer 1 accumulates packets with a predetermined initial delay amount and sequentially outputs temporarily accumulated packets to absorb fluctuations in packet arrival delays, ie jitter, so that packets can be output at a smoothed timing. Here, the output timing from the jitter absorbing buffer 1 follows an instruction from the output buffer monitoring unit 6 .

Packets output from the jitter absorbing buffer 1 are divided into background noise packets including background noise data and audio packets including audio coded data, and processed. In the case of an audio packet, the packet is input to the audio decoding unit 3 , and in the case of a background noise packet, the packet is input to the background noise generation unit 2 . Together with the background noise packet, the time difference between the background noise packet and the next packet, for example, the difference in the time stamp value indicating the transmission time given to the background noise packet and the next packet, respectively, is taken from the jitter absorption buffer as the background noise generation time length 1 is passed to the background noise generator 2.

Detailed operations are explained using diagrams. FIG. 2 is an explanatory diagram showing the relationship between the time stamp of a packet and accumulation in a jitter absorbing buffer.

In Figure 2, the audio packets #1, #2, #4 containing the audio coded data at time t, and the background noise packet #3 containing the background noise data are in the order of #1, #2, #3, #4 respectively arrives and is temporarily stored in the jitter absorbing buffer 1.

When assigning serial number N and time stamp value M to the grouping of #3 as the background noise grouping, the serial number of #1 grouping is N-2, the serial number of #2 grouping is N-1, and the serial number of #4 grouping is N+1, # The time stamp value of the 1 packet is M-2t, and the time stamp value of the #2 packet is M-t. The time stamp value of the #4 packet is the time after T time which is the length of the noise interval, that is, M+T. The background noise generation time length is the difference between the timestamp values of the #3 packet which is the background noise packet and the #4 packet which is the next packet, that is, (M+T)−M=T.

The background noise generation unit 2 to which the background noise group and the background noise generation time length are input generates background noise based on the background noise data stored in the background noise group, makes the generation of the background noise continue for the background noise generation time length, and serves as the background noise Audio data is output to the output buffer 5 .

The audio decoding unit 3 to which the audio packets are input decodes the encoded audio data stored in the audio packets to generate voice audio data, and outputs the audio data to the speech rate conversion unit 4 . The speech audio data processed by the speech rate conversion unit 4 is input to the output buffer 5 .

The output buffer monitoring unit 6 monitors the presence or absence of audio data stored in the output buffer 5 (accumulation amount of the stored audio data), and when it is determined that there is no input from the background noise generating unit 2 and the speech rate converting unit 4 (compared to When the predetermined amount is small), the jitter absorbing buffer 1 is instructed to output the timing of the packet so that one packet stored in the jitter absorbing buffer 1 is output.

The buffer remaining amount monitoring unit 71 monitors the amount of packets temporarily stored in the jitter absorbing buffer 1, and when the buffer remaining amount is less than a certain threshold value A, notifies the control signal output unit 72 of "small", and then When it exceeds a certain threshold value B, "large" is notified to the control signal output unit 72, and when it is above a certain threshold value A and below a certain threshold value B, "medium" is notified to the control signal output unit 72.

The control signal output unit 72 that has received the notification from the buffer remaining amount monitoring unit 71 outputs a time length control signal (instruction) that is controlled so that the buffer remaining amount of the jitter absorbing buffer 1 becomes higher and higher, and a playback speed control signal. The larger the background noise generation time length is, the shorter the reproduction speed control signal (instruction) is controlled so that the larger the buffer margin of the jitter absorbing buffer 1 is, the faster the speech speed is reproduced.

For example, according to the control content recorded in Table 1, if "small" is notified, an instruction to extend the background noise generation time length, for example, an instruction to extend by 1.1 times, is sent to the background noise generation part 2, and is sent to the speech rate conversion part 4. An instruction to play back slowly, for example, an instruction to slow down by a factor of 0.8. If "big" is notified, an instruction to shorten the background noise generation time length, for example, to 0.9 times, is sent to the background noise generation part 2, and an instruction to speed up the reproduction is sent to the speech rate conversion part 4, for example, to be 1.2 times faster instructions. If "medium" is notified, an instruction to set the background noise generation time length to a normal length, for example, 1.0 times, is sent to the background noise generation part 2, and an instruction to set the reproduction speed to a normal speed is sent to the speech rate conversion part 4. The indication is, for example, an indication of 1.0 times.

【Table 1】

As described above, according to the present embodiment, the control unit 7 issues an instruction to cooperate with the background noise generation unit 2 and the speech rate conversion unit 4 . That is, the duration of the background noise generated by the background noise generating unit 2 is controlled, and the reproduction speed converted by the speech rate converting unit 4 is controlled according to the accumulation status of packets in the jitter absorbing buffer 1 . In this way, background noise (silent interval) and speech (voiced interval) at different transmission intervals are separately controlled, so that voice quality degradation can be prevented even when applied to silent compression that does not always send packets at fixed intervals.

As the accumulation status of packets in the jitter absorbing buffer 1, according to the remaining capacity of the jitter absorbing buffer 1, a time length control signal for controlling the time length of the background noise generated by the background noise generating section 2, and a control signal converted by the speech rate are output. By using the reproduction speed control signal of the reproduction speed converted by the unit 4, appropriate jitter buffering control can be performed according to the margin of the jitter absorption buffer 1, and deterioration of call quality can be prevented even when silent compression is applied.

The jitter-absorbing buffer margin is described as "small", "medium", and "large" according to threshold A and threshold B, but finer control can be performed by subdividing the control.

In addition, although the control also changes with the change of the margin, by setting different thresholds for the thresholds of "small", "medium" and "large" according to the change direction of the margin, it is possible to avoid the Changes frequently due to the increase or decrease of the margin, which can provide better call quality. For example, it is possible to set the thresholds C and D in the case of a change direction in which the jitter absorption buffer margin increases, and the thresholds E, D in the case of a change in a decrease direction. Threshold F provides better call quality.

In addition, in the background noise generation unit 2 , it is possible to provide better call quality by making the background noise generation time not shorter than a certain fixed time length while shortening the background noise generation time length.

In addition, in the above description, the instruction from the control unit 7 to the background noise generating unit 2 is described as being extended by 1.1 times or shortened by 0.9 times, but for example, it may be extended by 100 ms or shortened by 200 ms. instructions.

In addition, although the case where the output buffer 5 and the output buffer monitoring part 6 were provided was demonstrated, the output buffer 5 and the output buffer monitoring part 6 may be deleted. For example, the jitter absorbing buffer 1 may be configured to output packets at output timings having predetermined time intervals. Furthermore, for example, the packet may be output at an output timing corresponding to the control of the control unit 7 in accordance with the accumulation state of the packet in the jitter absorbing buffer.

Implementation mode 2.

FIG. 3 is a functional block diagram showing an audio decoding device according to an embodiment of the present invention.

In FIG. 3 , the parts that are the same as or corresponding to those in the above-described embodiment are denoted by the same reference numerals, and description thereof will be omitted.

In FIG. 3 , the control unit 7 has a buffer remaining amount monitoring unit 71 , a control signal output unit 72 , and an arrival speed monitoring unit 73 . The arrival rate monitoring unit 73 monitors the arrival rate of packets stored in the jitter absorbing buffer 1 . In the present embodiment, the control signal output unit 72, based on the remaining amount monitored by the buffer remaining amount monitoring unit 71 and the arrival speed monitored by the arrival speed monitoring unit 73 as the packet accumulation status in the jitter absorbing buffer, A time length control signal for controlling the time length of the background noise generated by the background noise generation unit 2 and a reproduction speed control signal for controlling the reproduction speed converted by the speech rate conversion unit 4 are output.

Next, the operation will be described.

In addition, in this embodiment, the operation in the case where the user and the user's call partner are in a voice call is described, but the present invention is not limited thereto.

First, when the user's counterpart speaks, the voice is encoded at the counterparty's side to form packets, and is received at the user's side through the network. In this way, when the user side receives a packet transmitted from the communication partner side, the jitter absorbing buffer 1 temporarily stores the received packet. The jitter absorbing buffer 1 accumulates packets with a predetermined initial delay amount and sequentially outputs temporarily accumulated packets to absorb fluctuations in packet arrival delays, ie jitter, so that packets can be output at a smoothed timing. Here, the output timing from the jitter absorbing buffer 1 follows an instruction from the output buffer monitoring unit 6 .

Packets output from the jitter absorbing buffer 1 are divided into background noise packets including background noise data and audio packets including audio coded data, and processed. In the case of an audio packet, the packet is input to the audio decoding unit 3 , and in the case of a background noise packet, the packet is input to the background noise generation unit 2 . From the jitter absorbing buffer 1, together with the background noise packet, the time difference between the background noise packet and the next packet, for example, the difference between the time stamp values respectively indicating the transmission time given to the background noise packet and the next packet is generated as background noise. The length of time is passed to the background noise generation unit 2 .

The background noise generation unit 2 to which the background noise group and the background noise generation time length are input generates background noise based on the background noise data stored in the background noise group, makes the generation of the background noise continue for the background noise generation time length, and serves as the background noise Audio data is output to buffer 5 .

The audio decoding unit 3 to which the audio packets are input decodes the encoded audio data stored in the audio packets to generate voice audio data, and outputs the audio data to the speech rate conversion unit 4 . The speech audio data processed by the speech rate conversion unit 4 is input to the output buffer 5 .

The output buffer monitoring unit 6 monitors the presence or absence of audio data stored in the output buffer 5 (accumulation amount of the stored audio data), and when it is determined that there is no input from the background noise generating unit 2 and the speech rate converting unit 4 (compared to When the predetermined amount is small), the jitter absorbing buffer 1 is instructed to output the timing of the packet so that one packet stored in the jitter absorbing buffer 1 is output.

The buffer remaining amount monitoring unit 71 monitors the amount of packets temporarily stored in the jitter absorbing buffer 1, and when the buffer remaining amount is less than a certain threshold value A, notifies the control signal output unit 72 of "small", and then When it exceeds a certain threshold value B, "large" is notified to the control signal output unit 72, and when it is above a certain threshold value A and below a certain threshold value B, "medium" is notified to the control signal output unit 72.

The arrival rate monitoring unit 73 monitors the arrival rate of packets input (arriving) into the jitter absorbing buffer 1, and when the rate of arrival is slower than a certain threshold value α, notifies the control signal output unit 72 of “low speed”, and If the input speed is faster than a certain threshold value β, it is notified to the control signal output unit 72 as "high speed", and when it is not lower than a certain threshold value α and not higher than a certain threshold value β, it is regarded as "medium speed". Speed" is notified to the control signal output unit 72.

The control signal output section 72 that has received notifications from the buffer remaining amount monitoring section 71 and the arrival speed monitoring section 73 outputs a time length control signal and a reproduction speed control signal; this time length control signal (instruction) controls the jitter absorption buffer 1 The larger the buffer margin, the shorter the background noise generation time length is, and the higher the arrival speed of the packet input (arriving) into the jitter absorbing buffer 1 is, the shorter the background noise generation time length is; the reproduction speed control signal (instruction) controls The larger the buffer capacity of the jitter absorbing buffer 1 is, the faster the speech speed is reproduced, and the faster the arrival speed of packets input (arriving) into the jitter absorbing buffer 1 is, the faster the speech speed is reproduced.

For example, based on the control content described in Table 2, an instruction is given to the background noise generation unit 2 and the speech rate conversion unit 4 . For the background noise generating unit 2, an instruction of, for example, 1.1 times is issued when it is set to "extend", an instruction is issued, for example, 1.3 times when it is set to "further extend", and an instruction is issued when it is set to "shorten". For example, an instruction of 0.9 times, for example, an instruction of 0.5 times is issued when it is set to "further shortening", and an instruction of, for example, 1.0 times is issued when it is set as "normal". For the speech rate conversion unit 4, an instruction of, for example, 0.8 times is issued when it is set to "slow", an instruction of, for example, 0.6 times is issued when it is set to "slower", and an instruction is issued when it is set to "speed up". For example, an instruction of 1.2 times, for example, an instruction of 1.4 times is issued when it is set to "Further Speed", and an instruction of, for example, 1.0 times is issued when it is set as "Normal".

【Table 2】

As described above, according to the present embodiment, the control unit 7 issues an instruction to cooperate with the background noise generation unit 2 and the speech rate conversion unit 4 . That is, the duration of the background noise generated by the background noise generating unit 2 is controlled according to the accumulation status of the packets in the jitter absorbing buffer 1, and the reproduction speed converted by the speech rate converting unit 4 is controlled, whereby the transmission intervals are respectively controlled. Different background noise (silent intervals) and speech (voiced intervals), thus preventing deterioration of call quality even when applied to silent compression that does not necessarily send out packets at regular intervals.

Outputs the time to control the duration of the background noise generated by the background noise generating section 2 based on the remaining amount of the jitter absorbing buffer 1 and the arrival speed to the jitter absorbing buffer 1, which are the status of packet accumulation in the jitter absorbing buffer 1 The length control signal and the reproduction speed control signal for controlling the reproduction speed converted by the speech rate conversion unit 4 can perform appropriate jitter buffering control according to the remaining capacity of the jitter absorption buffer 1, and even if the reception of packets temporarily stops , and then when the stagnation is resolved and a large number of packets arrive at once, it is possible to realize appropriate jitter buffer control that can prevent buffer overflow by monitoring the arrival rate, and prevent deterioration of call quality even when silent compression is applied .

Although the jitter absorption buffer margin is classified into "small", "medium" and "large" according to threshold A and threshold B, and the arrival speed is classified into "low speed", "medium speed", The three types of "high speed" are explained, but finer control can be performed by further subdividing the control.

In addition, although the control also changes with the change of the jitter absorption buffer margin and the arrival speed, by distinguishing "small", "medium", "large", "low speed", "medium speed" according to the change direction of the margin and speed Setting different thresholds for the thresholds of "speed" and "high speed" can avoid frequent changes in the control due to the increase or decrease of the margin near the threshold, and can provide better call quality. For example, threshold C and threshold D in the case of a change direction in which the jitter absorption buffer margin increases, and threshold E and threshold F in the case of a change in a direction of decrease are set. . In addition, thresholds γ and δ are set when the direction of change increases toward an increase in arrival speed, and thresholds ε and ζ are set when the direction of change changes toward a direction of slowing down. As a result, better call quality can be provided.

In addition, in the background noise generation unit 2 , it is possible to provide better call quality by making the background noise generation time not shorter than a certain fixed time length while shortening the background noise generation time length.

In addition, in the above description, the instruction from the control unit 7 to the background noise generating unit 2 is described as 1.1 times or 0.9 times, but it may be an instruction about the increase or decrease in time, such as extending by 100 ms or shortening by 200 ms.

In addition, the controller 7 having the buffer remaining amount monitoring unit 71 and the arrival speed monitoring unit 73 has been described, but it may be configured such that the buffer remaining amount monitoring unit 71 is deleted, and the control signal output unit 72 is monitored by the arrival speed monitoring unit 73. The arrival speed of the arrival jitter absorbing buffer is outputted, and a time length control signal and a reproduction speed control signal are output.

In addition, although the case where the output buffer 5 and the output buffer monitoring part 6 were provided was demonstrated, the output buffer 5 and the output buffer monitoring part 6 may be deleted. For example, the jitter absorbing buffer 1 may be configured to output packets at output timings having predetermined time intervals. Furthermore, for example, the packet may be output at an output timing corresponding to the control of the control unit 7 in accordance with the accumulation state of the packet in the jitter absorbing buffer.

Implementation mode 3.

FIG. 4 is a functional block diagram showing an audio decoding device according to an embodiment of the present invention.

In FIG. 4 , the same reference numerals are used to designate the same or corresponding parts as those in the above-mentioned embodiment, and description thereof will be omitted.

In FIG. 4 , the high-precision silent compression unit 8 analyzes the received packet, and when a silent/noise interval is detected from the audio coded data included in the packet, replaces the packet with an interval containing background noise data. Background noise packet, in the case where no silence/noise interval is detected, the packet is output without permutation.

Next, the operation will be described.

In addition, in the present embodiment, the operation in the case of a voice call between the user and the user's counterpart is described, but the present invention is not limited thereto.

First, when the user's counterpart speaks, the voice is encoded at the counterparty's side to form packets, and is received at the user's side through the network. Silent compression is carried out in the coding on the counterparty side, and the background noise packet is output in the background noise section, and the audio packet is output in the voice section, and reaches the audio decoding device on the user side. When the accuracy of the silence compression function of the audio encoding device on the other party's side is poor, the packet is output as an audio packet regardless of whether it is actually a background noise section or not. Alternatively, the audio coding device on the other party's side does not implement the silent compression function, and outputs all packets as audio packets. In any case, the high-precision silent compression unit 8 is provided so that appropriate jitter absorption buffer control can be realized in the audio decoding device on the user side.

When the user side receives a packet transmitted from the call partner, the high-precision silent compression unit 8 analyzes the received packet, and finds a noise interval with higher precision from encoded data stored in the received audio packet. When a silent/noise interval is detected from the audio coded data included in the packet, the packet is replaced with a background noise packet including background noise data, and is output to the jitter absorbing buffer 1 . When no silent/noise interval is detected, the packet is output to the jitter absorbing buffer 1 without performing packet permutation. Subsequent operations are the same as in the above-mentioned embodiment.

As described above, according to this embodiment, a received packet is analyzed, and when a silence/noise interval is detected in the audio coded data included in the packet, the packet is replaced with a background noise packet including background noise data. , when the silence/noise interval is not detected, the packet is output without packet permutation, thereby controlling the background noise (silence interval) and voice (voiced interval), so appropriate jitter absorption buffer control can be realized, which can further prevent deterioration of call quality.

In addition, in this embodiment, the case where the arrival rate monitoring unit 73 monitors the arrival rate of the packet input to the high-precision silent compression unit 8 has been described, but it may be configured such that the high-precision silent compression unit 8 and the jitter absorption buffer 1 Monitor the arrival rate of packets.

In addition, the controller 7 having the buffer remaining amount monitoring unit 71 and the arrival speed monitoring unit 73 has been described, but it may be configured to have either one of the buffer remaining amount monitoring unit 71 and the arrival speed monitoring unit 73, and output the time Length control signal and reproduction speed control signal.

In addition, although the case where the output buffer 5 and the output buffer monitoring part 6 were provided was demonstrated, the output buffer 5 and the output buffer monitoring part 6 may be deleted. For example, the jitter absorbing buffer 1 may be configured to output packets at output timings having predetermined time intervals. Furthermore, for example, the packet may be output at an output timing corresponding to the control of the control unit 7 in accordance with the accumulation state of the packet in the jitter absorbing buffer.

Implementation mode 4.

FIG. 5 is a functional block diagram showing an audio decoding device according to an embodiment of the present invention.

In FIG. 5 , the same reference numerals are used to designate the same or corresponding parts as those in the above-mentioned embodiment, and description thereof will be omitted.

In FIG. 5, an audio decoding device 20 decodes audio coded data received at the user side. The audio encoding means 21 encodes speech to be transmitted from the user side. The audio detection unit 9 detects the presence or absence of a user's voice. In the present embodiment, whether the input audio data is "voice" or "noise" that is not voice is determined every fixed interval. When the audio data is "speech", it is determined that there is an utterance by the user, and when the audio data is "noise", it is determined that there is no utterance by the user.

The audio encoding unit 10 encodes audio data, and outputs encoded audio data. The silent compression control unit 11 outputs the audio coded data from the audio coding unit 10 when the audio detection unit 9 judges it as “speech”, and outputs the background noise from the audio coding unit 10 intermittently when it is judged as “noise”. data.

In addition, in the present embodiment, the jitter absorbing buffer 1 is configured to return the inside of the buffer to the initial state when the audio detection unit 9 detects the user's voice.

Next, the operation will be described.

In addition, in the present embodiment, the operation in the case of a voice call between the user and the user's counterpart is described, but the present invention is not limited thereto.

In the audio encoding device 21 , audio data is input to the audio detection unit 9 and the audio encoding unit 10 . The audio detection section 9 judges whether the input audio data is "speech" or "noise" that is not speech at regular intervals, and outputs the result to the audio coding section 10, the silent compression control section 11, and the audio data in the audio system. The jitter absorbing buffer 1 in the decoding device 20 . The audio coding unit 10 outputs coded data of the input audio data when it is notified that it is "voice", and outputs data of background noise when it is notified that it is "noise". The silent compression control unit 11 outputs the encoded audio data from the audio encoding unit 10 when it is notified that it is “speech”, and intermittently outputs the background noise data from the audio encoding unit 10 when it is notified that it is “noise”. . The determination result of the audio detection unit 9 is also notified to the jitter absorption buffer 1 . The jitter absorbing buffer 1 continues normal processing when it is notified that it is "noise", but when it is notified that it is "voice", it discards the audio packets stored in the jitter absorbing buffer 1, and starts from the initial state restarts processing.

When the audio data of "speech" is input to the audio coding device 21, the user is speaking, and normally, the user's other party is not speaking at this time. Therefore, in this case, there is a high possibility that there is no need to perform decoding processing on the user side, so by discarding the audio packets stored in the jitter absorbing buffer 1 and returning to the initial state, it is possible for the user's counterpart to start speaking and the user side When starting the decoding process, it is possible to perform jitter absorbing buffer control from an initial state that is not a state close to buffer exhaustion or overflow.

As described above, according to the present embodiment, when the audio data of "speech" is input to the audio coding device 21, the audio packets stored in the jitter absorbing buffer 1 are discarded and returned to the initial state, and the user's communication partner When utterance is started and decoding processing is started on the user side, the jitter absorbing buffer control can be performed from an initial state not close to the buffer exhaustion or overflow state, so that more appropriate control can be realized, and deterioration of call quality can be further prevented.

In addition, the audio encoding device 21 does not necessarily need to apply the silent compression, and the audio detection unit 9 may be provided, and the jitter absorption buffer 1 may obtain the determination result.

In addition, the controller 7 having the buffer remaining amount monitoring unit 71 and the arrival speed monitoring unit 73 has been described, but it may be configured to have either one of the buffer remaining amount monitoring unit 71 and the arrival speed monitoring unit 73, and output the time Length control signal and reproduction speed control signal.

In addition, although the case where the output buffer 5 and the output buffer monitoring part 6 were provided was demonstrated, the output buffer 5 and the output buffer monitoring part 6 may be deleted. For example, the jitter absorbing buffer 1 may be configured to output packets at output timings having predetermined time intervals. Furthermore, for example, the packet may be output at an output timing corresponding to the control of the control unit 7 in accordance with the accumulation state of the packet in the jitter absorbing buffer.

Implementation mode 5.

FIG. 6 is a functional block diagram showing an audio decoding device according to an embodiment of the present invention.

In FIG. 6 , the same or corresponding parts as those in the above-described embodiment are denoted by the same reference numerals, and description thereof will be omitted.

In Fig. 6, the background noise data detection/insertion section 12 detects whether the received packet contains background noise data, and if it detects that the background noise data is included, compares the number with the time length of the silent/noise interval of the background noise data Corresponding packets whose time length per packet is equal to the time length of each packet including audio coded data are inserted into the jitter absorbing buffer 1 .

Next, the operation will be described.

In addition, in the present embodiment, the operation in the case of a voice call between the user and the user's counterpart is described, but the present invention is not limited thereto.

First, when the user's counterpart speaks, the voice is encoded at the counterparty's side to form packets, and is received at the user's side through the network.

In the background noise data detection/insertion section 12, it is detected whether the received packet is a background noise packet containing background noise data, and if a background noise packet is detected, the number is compared with the number of the silent/noise interval of the background noise data. Packets having an equivalent time length are inserted into the jitter absorbing buffer 1 , and the time length of each of these packets is equal to the time length of each packet including audio coded data.

Detailed operations will be described using drawings. FIG. 7 is an explanatory diagram showing the relationship between the time stamp of a packet and accumulation in a jitter absorbing buffer.

In FIG. 7, audio packets #1, #2, #4 containing audio coded data at time t and background noise packet #3 containing background noise data arrive in the order of #1, #2, #3, #4 and It is temporarily stored in the jitter absorption buffer 1. When assigning serial number N and time stamp value M to the grouping of #3 as the background noise grouping, the serial number of #1 grouping is N-2, the serial number of #2 grouping is N-1, and the serial number of #4 grouping is N+1, # The time stamp value of the 1 packet is M-2t, and the time stamp value of the #2 packet is M-t. The time stamp value of the #4 packet is the time after T time which is the length of the noise interval, that is, M+T.

The background noise data detecting/inserting section 12, upon detecting the #3 packet as the background noise packet, stores its sequence number N and time stamp value M in advance, outputs the #3 packet to the jitter absorbing buffer 1, and waits for the #3 packet to be used as the next packet The arrival of the packet whose sequence number is N+1. The background noise data detection/insertion unit 12 finds the time stamp value M+T when the packet with the sequence number N+1, that is, the #4 packet arrives, and calculates the time length T of the noise interval existing between the #2 packet and the #4 packet. In order to make background noise packets exist at intervals of time t like audio packets existing at intervals of t, X background noise packets at time t corresponding to the noise interval of time T are inserted into # of the jitter absorbing buffer 1 After the 2 packet, the #4 packet is then output to the jitter absorbing buffer 1. As a result, audio packets or background noise packets exist in the jitter absorbing buffer 1 at intervals of t.

The buffer remaining amount monitoring unit 71 monitors the amount of packets temporarily stored in the jitter absorbing buffer 1, and when the buffer remaining amount is less than a certain threshold value A, notifies the control signal output unit 72 of "small", When it exceeds a certain threshold value B, "large" is notified to the control signal output unit 72, and when it is above a certain threshold value A and below a certain threshold value B, "medium" is notified to the control signal output unit 72 .

The arrival speed monitoring unit 73 monitors the arrival speed of the packet input (arriving) into the jitter absorbing buffer 1, and notifies the control signal output unit 72 of the “low speed” when the arrival speed of the packet is slower than a certain threshold value α. If the input is performed at a speed faster than a certain threshold β, it is notified to the control signal output unit 72 as "high speed", and when it is not lower than a certain threshold α and not higher than a certain threshold β, it is regarded as "high speed". “Medium speed” is notified to the control signal output unit 72 .

The control signal output unit 72 outputting (instructing) the control signal output unit 72 receiving the notification from the buffer remaining capacity monitoring unit 71 and the arrival speed monitoring unit 73 is controlled so that the larger the buffer capacity of the jitter absorbing buffer 1, the shorter the background noise generation time length, input (Arrival) The faster the arrival speed of the packet in the jitter absorbing buffer 1 is, the shorter the background noise generation time length control signal is, and the output (indication) is controlled so that the larger the buffer margin of the jitter absorbing buffer 1 is, the faster the speech rate is. A reproduction speed control signal for reproducing and reproducing a packet input (arrived) into the jitter absorbing buffer 1 at a faster speech rate as the arrival speed is higher.

The control signal output unit 72 that receives the notification from the buffer remaining amount monitoring unit 71 and the speed of arrival monitoring unit 73 issues instructions to the jitter absorbing buffer 1 and the speech rate conversion unit 4 based on the control content described in Table 2, for example. . For the jitter absorbing buffer 1, for example, an instruction to insert 1 background noise packet is issued in the case of "extending", an instruction is issued for example to insert 3 background noise packets in the case of "further extension", and in the case of "shortening" In the case of ", for example, an instruction to delete one background noise packet is issued, in the case of "further shortening", an instruction to delete, for example, three background noise packets is issued, and in the case of "normal", an instruction is issued, for example, no insertion/deletion instruct. For the speech rate conversion unit 4, an instruction of, for example, 0.8 times is issued in the case of "slow", an instruction of, for example, 0.6 times is issued in the case of "slower", and an instruction of, for example, 1.2 times is issued in the case of "faster". In the case of "further acceleration", for example, an instruction of 1.4 times is issued, and in the case of "normal", an instruction of, for example, 1.0 times is issued.

As described above, according to the present embodiment, the control unit 7 issues a linkage instruction to the jitter absorbing buffer 1 and the speech rate converting unit 4 based on the remaining amount of the jitter absorbing buffer and the arrival speed. That is, the duration of the background noise generated by the background noise generating unit 2 is controlled according to the accumulation status of the packets in the jitter absorbing buffer 1, and the reproduction speed converted by the speech rate converting unit 4 is controlled, whereby the transmission intervals are respectively controlled. Different background noise (silent intervals) and speech (voiced intervals), thus preventing deterioration of call quality even when applied to silent compression that does not necessarily send out packets at regular intervals.

When it is detected that background noise data is included, the number of packets corresponding to the time length of the silence/noise interval of the background noise data is inserted into the jitter absorbing buffer 1 to control the background noise generated by the background noise generating unit 2 The time length of each packet is equal to the time length of each packet containing the audio coded data, so that the number of packets accumulated in the jitter absorbing buffer 1 can be Since the control is performed, the processing of the background noise generation unit 2 can be simplified.

Also, even when packet reception is temporarily stagnant, and the stagnation is resolved and a large number of packets arrive at once, appropriate jitter buffer control capable of preventing buffer overflow from occurring can be realized by monitoring the arrival rate.

According to threshold A and threshold B, the jitter absorption buffer margin is divided into three categories: "small", "medium", and "large", and according to threshold α and threshold β, the arrival speed is divided into "low speed", "medium speed", " The three categories of "high speed" are described, but more detailed control can be carried out by further subdividing the control.

In addition, the control also changes with the change of the jitter absorption buffer margin and the arrival speed, but by distinguishing "small", "medium", "large", "low speed", "medium speed" according to the change direction of the margin and speed ” and “High Speed” thresholds can be set with different thresholds, which can avoid frequent changes in the control due to the increase or decrease of the margin near the thresholds, and can provide better call quality. For example, threshold C and threshold D in the case of a change direction in which the jitter absorption buffer margin increases, and threshold E and threshold F in the case of a change in a direction of decrease are set. . In addition, thresholds γ and δ are set when the direction of change increases toward an increase in arrival speed, and thresholds ε and ζ are set when the direction of change changes toward a direction of slowing down. As a result, better call quality can be provided.

In addition, in this embodiment, the description is based on the packetization period, but when a plurality of audio coded frames are included in one packet, control may be performed based on the time length of the audio coded frame.

In addition, as an operation of the background noise data detection/insertion unit 12, the background noise packets may be sequentially inserted into the jitter data every time t after the arrival of the #3 packet as the background noise packet and before the #4 packet as the audio packet. Absorption buffer 1 inside.

In addition, in the background noise generation unit 2 , when the background noise generation time length is shortened so that the background noise generation time does not become shorter than a certain fixed time length, better call quality can be provided.

In addition, the controller 7 having the buffer remaining amount monitoring unit 71 and the arrival speed monitoring unit 73 has been described, but the arrival speed monitoring unit 73 may be deleted, and the time length may be output according to the monitoring result of the buffer remaining amount monitoring unit 71. control signal and reproduction speed control signal.

In addition, although the case where the output buffer 5 and the output buffer monitoring part 6 were provided was demonstrated, the output buffer 5 and the output buffer monitoring part 6 may be deleted. For example, the jitter absorbing buffer 1 may be configured to output packets at output timings having predetermined time intervals. Furthermore, for example, the packet may be output at an output timing corresponding to the control of the control unit 7 in accordance with the accumulation state of the packet in the jitter absorbing buffer.

Claims (7)

1. An audio decoding device, characterized in that it has: a jitter absorbing buffer temporarily accumulating received packets and outputting the packets at a prescribed output timing; a background noise generating unit that generates audio data of background noise based on the background noise data included in the packet output from the jitter absorbing buffer; an audio decoding unit that decodes encoded audio data included in the packet output from the jitter absorbing buffer to generate audio data of speech; a speech rate conversion unit that performs speech rate conversion for converting the reproduction speed of the audio data decoded by the audio decoding unit; and The control unit controls the duration of the background noise generated by the background noise generation unit and the playback speed converted by the speech rate conversion unit based on the accumulation status of packets in the jitter absorption buffer. 2. audio decoding device according to claim 1, is characterized in that, The above-mentioned control unit has: a buffer remaining amount monitoring unit that monitors the remaining amount of the jitter absorbing buffer as the storage status; and a control signal output unit for outputting a time length control signal for controlling the time length of the background noise generated by the background noise generation unit, and a control signal for controlling the speech rate by the above-mentioned speech rate, based on the remaining capacity monitored by the buffer remaining capacity monitoring unit. The reproduction speed control signal of the reproduction speed converted by the conversion unit. 3. audio decoding device according to claim 1, is characterized in that, The above-mentioned control unit has: an arrival rate monitoring unit that monitors the arrival rate of the received packet at the jitter absorbing buffer as the storage status; and a control signal output unit for outputting a time length control signal for controlling the time length of the background noise generated by the background noise generation unit based on the arrival speed monitored by the arrival speed monitoring unit; The converted reproduction speed control signal of the above reproduction speed. 4. audio decoding device according to claim 1, is characterized in that, The audio decoding device includes a high-precision silent compression unit, and the high-precision silent compression unit analyzes the received packet, and when a silent/noise interval is detected from encoded audio data contained in the packet, the packet is divided into replacing with a background noise packet containing background noise data, outputting the above packet without replacement if the above-mentioned silence/noise interval is not detected, The jitter absorbing buffer temporarily stores packets output from the high-precision silent compression unit. 5. audio decoding device according to claim 1, is characterized in that, The above-mentioned audio decoding device has an audio detection unit that detects the presence or absence of a user's voice, The jitter absorbing buffer returns to an initial state when the audio detection unit detects a user's utterance. 6. audio decoding device according to claim 1, is characterized in that, The audio decoding device has a background noise data detection/insertion unit, the background noise data detection/insertion unit detects whether or not the received packet contains background noise data, and compares the number with the background noise data if background noise data is detected. Insert the following packets into the above-mentioned jitter absorbing buffer whose time length is equivalent to the silence/noise period of the noise data, and the time length of each packet of this packet is equal to the time length of each packet including the audio coded data packet . 7. audio decoding device according to claim 1, is characterized in that, has: an output buffer temporarily accumulating audio data of the aforementioned background noise and audio data of the aforementioned speech; and an output buffer monitoring unit that monitors an accumulation amount of the audio data accumulated in the output buffer, and instructs the jitter absorbing buffer to output timing of the temporarily accumulated packets based on the accumulation amount, The jitter absorbing buffer outputs the temporarily stored packets according to an instruction from the output buffer monitoring unit.