JP2000307654A - Voice packet transmitting system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To correct the discontinuity of the levels of a silence interval and a sound interval without damaging the effective use of a transmission line in silence control by inserting an interpolating frame generated by an interpolating frame generating means at the boundary between the sound interval and the silence interval of a voice frame. SOLUTION: An interpolating frame generated by an interpolating frame generating means is inserted to the boundary between the sound interval and the silence interval of a voice frame based on the judging result of a silence boundary judging means for judging the boundary between the sound interval and the silence interval of a received voice frame. In this system, a voice frame string stored in a voice frame buffer 303 is subjected to silence control detection by a silence control frame detecting part 304, and a background noise frame generated by a background noise frame generating part 111 and the interpolating frame interpolated by a voice frame interpolating part 305 are inserted and the resultant string is converted to the voice frame string. Subsequently, the string is sent to a voice extending part 108 to restore a voice signal.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice packet transmission system for transmitting a voice signal via a packet network, and more particularly to a voice packet transmission system for performing a silent control that does not transmit a packet in a silent state.

[0002]

2. Description of the Related Art Conventionally, a voice packet transmission system for transmitting a voice signal via a packet network and performing no-speech control without transmitting a packet in a no-sound state, detects a voiced section / silent section on the transmitting side. , The packet is not transmitted in the silent section, the background noise is reproduced on the receiving side, and so on.
The transmission line is effectively used.

However, the sound pressure level at the boundary between a silent section and a sound section changes continuously on the input side, but changes discontinuously on the output side. There is a problem of feeling uncomfortable.

[0004] As disclosed in Japanese Patent Application Laid-Open No. 5-292121, a method for initializing differential encoding for a break or noise at a boundary between a silent section and a voiced section is disclosed.
As disclosed in JP-A-116571, a method of outputting a prediction coefficient as the first packet of a voiced section and matching the states of the encoders, a silent section in front of a voiced section for delaying voice is described as a voiced section. A method of transmitting a voice frame and transmitting the voice frame has been proposed.

[0005]

However, Japanese Patent Laid-Open Publication No.
The techniques described in JP-A-92121 and JP-A-9-116571 are means for correcting an error of a decoder due to silent packet discarding. Cannot be corrected.

[0006] The method of treating as silence from a silence section ahead of a speech section is negative in terms of effective use of a transmission path by silence compression. Further, there is a problem that a sense of incongruity in silence control remains due to the discontinuity between the pseudo background noise and the speech frame of the silence section actually transmitted.

According to the present invention, in a silent control, a voice packet which does not cause a sense of incongruity at the beginning or end of a conversation by correcting the level discontinuity between a silent section and a sound section without impairing the effective use of a transmission path. It is intended to provide a transmission system.

[0008]

SUMMARY OF THE INVENTION The present invention provides a voice encoding means for converting a voice signal into a digitized voice frame, a voice decoding means for converting the voice frame into a voice signal, and packetizing the voice frame. Packet transmitting and receiving means for transmitting and receiving, silence detecting means for detecting the presence or absence of voice based on the audio signal, silence control means for controlling the transmission and reception of voice frames by the packet transmitting and receiving means based on the detection result of the silence detecting means, A silent boundary determining means for determining a boundary between a voiced section and a silent section of the received voice frame; and an interpolation frame generating means for generating at least one interpolation frame for compensating for a gap between voice frames. Based on the determination result of the means, the interpolation frame generating means is provided at the boundary between the sound section and the silent section of the voice frame. It is intended to operate to insert the at least one interpolated frame generated me.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS [First Embodiment] FIG. 1 is a block diagram schematically showing a packet transmission system according to one embodiment of the present invention.

In the above embodiment, an audio input unit 101 for inputting an analog audio signal from a microphone, a telephone line or the like, and an A for converting an analog audio signal into a PCM digital audio signal.
A / D converter 102, an audio compressor 103 for converting a PCM digital audio signal into an audio frame obtained by compression encoding,
A packet transmission unit 104 for assembling voice frames into packets that can be transferred on a packet network; a silence detection unit 105 for detecting a silent section of a voice signal;
0, a packet transmitting / receiving interface 106 for transmitting / receiving packets, a packet receiving unit 107 for decomposing a voice frame from a received packet, and a background noise frame generating unit 111 for generating a background noise frame.
A voice decompression unit 108 for decompressing a compressed voice frame and converting it into a PCM digital voice signal;
D / A for converting digital audio signals to analog audio signals
It has a conversion unit 109 and an audio output unit 110 that outputs an analog audio signal to a speaker, a telephone line, or the like.

FIG. 2 is a block diagram showing the configuration of the packet transmission unit 104.

[0012] The packet transmission unit 104 includes an audio frame buffer 201 for temporarily storing the audio frame converted by the audio compression unit, and a silence control frame generation unit for generating a frame for performing silence control based on the detection result from the silence detection unit 105. A packet assembler 202 that acquires a voice frame necessary for packetization from a voice frame buffer, adds header information such as a time stamp, a sequence number, and a destination address to assemble the packet;
The assembled packet is sent to the packet interface (I /
And F) a packet transmission buffer 203 to be transferred to 106.

FIG. 3 is a block diagram showing the configuration of the packet receiving unit 107.

A packet receiving unit 107 stores a packet received from the packet interface 106, and based on the header information, a packet receiving buffer 301 for controlling packet rearrangement and processing timing, and removes the header information of the packet to remove the audio frame. The packet decomposing unit 302 that cuts out the audio frame and the audio frame buffer 303 that stores the audio frame detect a silence control frame, acquire the background noise frame from the background noise frame generation unit 111, and store the silence control frame in the audio frame buffer 303. And an audio frame interpolation unit 305 for predicting and calculating an arbitrarily located audio frame from the arrangement of audio frames stored in the audio frame buffer.

Next, the operation of the above embodiment will be described.

FIG. 4 is a diagram showing the audio signal 401.

The audio signal 401 is converted into an audio frame sequence 402 by the audio compression section 103 as shown in FIG. As an audio compression method, ITU-T Recommendation G. 7
28, G. 729, G.C. For example, a hybrid system specified in 723.1 and the like can be considered.

FIG. 6 is a flowchart showing an encoding processing cycle in a sound section in the above embodiment.

In the flow chart of FIG. 6, since the audio frame subjected to the audio compression processing (S601) by the audio compression unit 103, for example, the audio frame F3 of the audio frame sequence 402, is not detected by the audio absence detection unit 105, no audio is detected (S602, No). ), The audio frame F3 is stored in the audio frame buffer 201 (S603).

Here, for example, an audio frame sequence 402
When the input audio signal 401 is detected to be lower than the reference value L1 and changed to silence as in the timing of the audio frame F8 (S602, Yes), the audio frame F8 is discarded (S611). The generation unit 204 stores a silent control on-frame Fon indicating a change from a sound section to a silent section in the audio frame buffer 201 (S612), and transits to a silent section (S613).

FIG. 7 is a flowchart showing a coding cycle process in a silent section in the above embodiment.

First, an audio frame subjected to audio compression processing by the audio compression unit 103 (S701), for example, an audio frame F12 of the audio frame sequence 402,
Since no sound is detected by the method No. 05 (S702, N
o), it is not stored in the audio frame buffer 201 and is discarded (S703).

Here, for example, the voice frame sequence 402
When it is detected that the input audio signal 401 has exceeded the reference value L1 and has changed to a voiced section (S702, Yes), as in the timing of the voice frame F18 in (S702, Yes), the voiceless The silence control off-frame Foff indicating the change to the sound section is stored in the audio frame buffer 201 (S711), and the audio frame F18 is continuously stored in the audio frame buffer 20.
1 (S712), and transits to a sound section (S7).
13).

The audio frame buffer 20 shown in FIG.
The sequence 403 of voice frames and silence control frames stored in No. 1 is assembled into a packet sequence 404 by the packet assembling unit 202 and transferred to the packet transmission buffer 203.

FIG. 8 is a flowchart showing the packet assembling cycle processing in the above embodiment.

First, when an audio frame exists in the audio frame buffer 201 (S801, Yes), an audio frame necessary for packet assembly is obtained from the audio frame buffer 201 (S802), and the obtained audio frame is synthesized. (S803), and adds header information (S803).
804) and store it in the packet transmission buffer 203 (S
805). If no audio frame exists in the audio frame buffer 201 (S801, No), the process ends.

FIG. 4 shows a case where the number of frames to be synthesized is 1, that is, the audio frame and the audio packet are in a one-to-one correspondence.

FIG. 9 is a flowchart showing the packet transfer cycle processing in the above embodiment.

FIG. 10 is a flowchart showing the receiving process in the above embodiment.

First, the packet stored in the packet transmission buffer 203 is taken out of the packet transmission buffer (S901), and if a packet exists (S902,
Yes), the packet is transmitted to the packet network 120 via the packet interface (I / F) 106 (S903).
If there is no packet (S902, No), the process ends. The packet transmitted to the packet network 120 is
When the data is transferred based on the header information and received via the packet I / F 106, it is stored in the packet reception buffer 301 in the reception processing shown in FIG. 10 (S100).
1).

As shown in FIG. 5, the received packet sequence 501 stored in the packet receiving buffer 301 is decomposed into a received frame sequence 502 by the packet decomposing unit 302 and stored in the voice frame buffer 303.

FIG. 11 is a flowchart showing a packet disassembly cycle in the above embodiment.

First, based on the sequence number information of the packet header, the inversion of the order of arrival of the received packet sequence 501 in the packet receiving buffer 301 is corrected (S1).
101). Next, based on the time stamp information of the packet header, it is determined whether or not there is a packet to be decomposed in the packet reception buffer 301 (S11).
02), when there is a packet at the processing timing (S
1102, Yes), the packet is extracted from the reception packet buffer 301 (S1103), the header is removed (S1104), and the frame is decomposed (S110).
5) The audio frame is stored in the audio frame buffer 303 (S1106). If there is no packet at the disassembly processing timing (S1102, No), the processing ends as it is.

As shown in FIG. 5, the speech frame sequence 502 stored in the speech frame buffer 303 is subjected to silence control detection by the silence control frame detection unit 304, and the background generated by the background noise frame generation unit 111. The noise frame Fs and the interpolated frame Fi interpolated by the audio frame interpolation unit 305 are inserted, converted into an audio frame sequence 503, and then sent to the audio expansion unit 108 to restore the audio signal 504.

FIG. 12 shows a speech section, for example, the speech frame F6 of the speech frame sequence 502 in the above embodiment.
9 is a flowchart showing a decoding cycle process at the timing of FIG.

First, it is determined whether or not a boundary between a silent section and a sound section is present based on a change between a speech frame and a background noise frame (S1201). In the present embodiment, if replacement processing by an interpolation frame described later is performed for one frame, the boundary is determined by three consecutive frames of a voiced frame and two background noise frames. The frames F4 to F6 are all sound frames, and are determined not to be boundaries (S1201, No).

Next, it is determined whether or not the frame is the silent control control frame Fon (S1202). Since the frame is not Fon (S1202, No), the first audio frame in the audio frame buffer 303 is extracted (S1203).
The data is transferred to the audio decompression unit 108 (S1204).

Here, since the audio frame buffer 303 stores at least the audio frames necessary for the silent boundary detection, the audio frames transferred to the audio decompression unit 108 are those before the frame F4. In the decoding cycle processing at the timing of the silent control on-frame Fon subsequent to the sound interval F7 in the voice frame sequence 502, the silent control on-frame is recognized (S120).
2, Yes), the state transits to the silent section (S121).
1).

In the decoding cycle processing at the timings of the sound sections F2 and F18, it is determined that the boundary is between the sound section and the sound section (S1201, Yes), and the sound is located in the sound section and the sound section. The background noise frames are replaced with interpolation frames Fi0 and Fi2, respectively (S12).
11). For example, in the case of the hybrid coding method, a method of generating an interpolation frame using a filter coefficient and a noise codebook index in a sound section is used, and a gain coefficient is set to an intermediate value with a background noise gain. .

FIG. 13 is a flowchart showing a decoding cycle process in a silent section in the above embodiment.

First, it is not determined that the boundary between the sound section and the silent section (S1301, No), and the silent control off frame F
If it is determined that it is not off (S1302, No),
The background noise frame Fs is stored in the audio frame buffer 303 (S1303), the first audio frame is extracted (S1304), and transferred to the audio expansion unit 108 (S1305). If it is determined that the boundary is a boundary between a sound section and a silent section (S1301, Yes), the interpolation frame F
i1 and Fi3 are generated and replaced (S1311), and when the silence control off-frame Foff is detected (S1302, Y
es), the state is transited to the sound section (S1321).

[Second Embodiment] In the first embodiment, the size of the interpolation frame is changed according to the temporal length of one audio frame determined by the audio compression method. , It is possible to keep the temporal change of the boundary with the same level, and to realize stable smoothness regardless of the audio compression method.

[0043]

According to the present invention, in the silence control, the level discontinuity between the silence section and the speech section is corrected without impairing the effective use of the transmission path, and a sense of discomfort is obtained at the beginning or end of a conversation. It is possible to build a voice packet transmission system that you can not feel, and furthermore, regardless of the voice compression method,
There is an effect that stable smoothness can be realized.

[Brief description of the drawings]

FIG. 1 is a block diagram schematically showing a packet transmission system according to one embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of a packet transmission unit 104.

FIG. 3 is a block diagram illustrating a configuration of a packet receiving unit 107.

FIG. 4 is a diagram showing an audio signal 401.

FIG. 5 is a diagram showing an audio signal 501.

FIG. 6 is a flowchart showing an encoding processing cycle in a sound section in the embodiment.

FIG. 7 is a flowchart showing an encoding cycle process in a silent section in the embodiment.

FIG. 8 is a flowchart showing a packet assembly cycle process in the embodiment.

FIG. 9 is a flowchart showing a packet transfer cycle process in the embodiment.

FIG. 10 is a flowchart showing a packet transfer cycle process in the embodiment.

FIG. 11 is a flowchart showing a packet disassembly cycle in the embodiment.

FIG. 12 is a flowchart showing a decoding cycle process of the timing of a sound section, for example, the audio frame F6 of the audio frame sequence 502 in the embodiment.

FIG. 13 is a flowchart showing a decoding cycle process in a silent section in the embodiment.

[Explanation of symbols]

 100: Packet transmission system, 104: Packet transmission unit, 105: Silence detection unit, 107: Packet reception unit, 111: Background noise frame generation unit, 120: Packet network, 401: Voice signal.

Claims (3)

[Claims] 1. An audio packet transmission system for transmitting an audio signal via a packet network, comprising: audio encoding means for converting the audio signal into a digitized audio frame; and converting the audio frame into an audio signal. Voice decoding means for packetizing the voice frame and transmitting and receiving the voice frame; and silence detecting means for detecting the presence or absence of voice based on the voice signal;
Silence control means for controlling transmission and reception of a voice frame by the packet transmission / reception means based on the detection result of the silence detection means; silence boundary determination means for determining a boundary between a sound section and a silence section in a received voice frame; Interpolated frame generation means for generating at least one frame that supplements between the audio frame sequences; and based on the determination result of the silent boundary determination means, at the boundary between the sound interval and the silent interval in the audio frame, A voice packet transmission system, wherein at least one interpolation frame generated by the interpolation frame generation means is inserted. 2. The apparatus according to claim 1, further comprising background noise frame generation means for generating a background noise frame, wherein the silence control means is generated by the background noise frame generation means when a received voice frame is in a silent state. Means for acquiring a background noise frame, wherein the audio decoding means converts the background noise frame into an audio signal, and converts the background noise frame existing at the boundary between the background noise frame and the audio frame into the interpolation frame. A voice packet transmission system, which is means for replacing the interpolation frame generated by the generation means. 3. The voice frame time recognition means according to claim 1, further comprising: a voice frame time recognition means for recognizing a time length of one voice frame; A voice packet transmission system, characterized in that it is means for determining the number of interpolated frames on the basis of the above.