JP2003241799A - Acoustic encoding method, decoding method, encoding device, decoding device, encoding program, decoding program - Google Patents

Abstract

(57) [Summary] [Problem] To provide an acoustic encoding method that applies scalable encoding and suppresses sound quality degradation even in a situation where packet loss occurs. A waveform is cut out at predetermined time intervals from an input audio signal (step 101), and scalable encoding is executed (step 102). A code bit string A from a first band (main layer A), a code bit string B from a second band (auxiliary layer B) including a coding error in the first band,
A code bit sequence C from a third band (auxiliary layer C) including a coding error up to the second band is obtained. These code bits are packetized so as to become an independent packet for each layer (step 103). If necessary, error protection is performed so that the priority on the first band side becomes higher (step 104), and the packet is transmitted (step 1).
05).

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to an acoustic coding method,
The present invention relates to a decoding method, an encoding device and a decoding device, and in particular,
The present invention relates to a scalable acoustic coding method and acoustic coding apparatus that receive an acoustic signal as an input, and a corresponding acoustic decoding method and acoustic decoding apparatus.

[0002]

2. Description of the Related Art In the coding of a musical tone signal or a voice signal, there has been no scalable coding technique in the past. Therefore, when a voice or musical tone code bit string is distributed via a network, it is distributed by non-scalable coding. Was. For non-scalable encoded delivery,
The code bit string to be delivered is 1 when the packet delivery is used.
Packetization has been performed by using one encoded frame as one packet, a plurality of encoded frames as one packet, or one encoded frame as a plurality of packets. In addition, methods have been devised to increase resistance to packet loss by interleaving packets or retransmitting packets when packet loss occurs. However, since these conventional methods are non-scalable coding schemes, the information contained in each packet is the information over the entire band of the signal sequence in which the input signal is divided in a predetermined time domain, and therefore a certain packet. Is lost, the sound of the entire band included in the packet is lost, and appears as large noise or distortion. That is, the transmission of the acoustic signal by non-scalable coding is not always preferable from the viewpoint of hearing.

Considering, for example, that a code bit string of voice or musical sound is packetized and delivered on the Internet or the like, when congestion occurs due to the concentration of traffic on the network or when the jitter absorption of a received packet fails, the packet is Sound quality deterioration due to loss occurs. This sound quality deterioration appears as sound interruptions and burst noises, and is very annoying. In particular, when transmitting audio signals such as voices and musical sounds using VoIP (Voice over IP), Internet radio, Internet TV, etc. on the Internet, it is important that live transmission is performed and packet reproduction processing is performed. Since it is difficult to adopt, such packet loss becomes a big problem, and if the deterioration of the sound due to the packet loss can be suppressed, the acoustic signal can be comfortably transmitted.

By the way, the inventors of the present invention have disclosed the patent No. 3139.
No. 602 (or Japanese Unexamined Patent Publication No. H8-263096) describes an audio signal of a certain band (first band) as an audio signal coding method by scalable coding that gives selectivity to decoding quality and coding compression rate. Has been proposed, and a method of repeating a series of processing of encoding the sound of the second band that is wider than the first band and the coding residual of the first band, and repeating the processing. According to this method, the lower layer (first band side)
Even when the encoding by different compression techniques is applied to the upper layer and the upper layer, the encoding quality does not deteriorate in the decoded signal up to the upper layer, and the decoding quality in the auditory sense does not occur at any layer. It has the effect of being optimal.

However, even with such a scalable coding, the conventional method may cause a considerable sound quality deterioration in the sense of hearing when there is a packet loss or the like.

[0006]

As described above, when an acoustic signal is coded by the non-scalable coding method and the coded signal is packetized, sound loss or noise may occur due to packet loss during transmission. There is a problem that burst noise appears. Even if the scalable coding method is used, sound quality deterioration may occur due to packet loss. Therefore, there is a demand for an encoding method and a packet transmission method that do not significantly deteriorate the sound quality even when a packet is lost.

An object of the present invention is to apply scalable coding and to suppress sound quality deterioration even in a situation where packet loss occurs, or to subject a transmission packet to error protection so as to suppress sound quality deterioration. Another object of the present invention is to provide an acoustic encoding method and an acoustic encoding device capable of controlling the strength of error protection, and an acoustic decoding method and an acoustic decoding device corresponding thereto.

[0008]

A first acoustic coding method of the present invention is an acoustic coding method for coding and transmitting an acoustic signal, wherein scalable coding is performed on an input acoustic code, The code obtained for each layer in the scalable coding is packetized, and the packet of each layer is distributed. Here, it is preferable that the packet of each layer is delivered after performing the error protection processing so that the packet of the layer of higher importance has a higher priority.

A second acoustic encoding method of the present invention is an acoustic encoding method for encoding an acoustic signal for every N (N is an integer of 2 or more) bands, where i is 2 or more and N or less. As each integer, the i-th band includes the (i-1) -th band, and the first extraction step of extracting the component signal of the first band of the acoustic signal and the first extraction process by encoding the component signal of the first band. The first encoding process for obtaining the first code, the (i-1) th decoding process for obtaining the decoded signal in the (i-1) th band based on the first to (i-1) th codes, and the i-th acoustic signal The i-th extraction step of extracting the component signal of the i-th band and the i-th extraction step of encoding the residual signal of the component signal of the i-th band and the decoded signal of the i-1-th band to obtain the i-th code The encoding process and the process of transmitting the first to Nth codes separately for each band are included. Where the i-th
It is preferable to perform the error protection processing so that the transmission error or omission in the -1 band is smaller than the transmission error or omission in the i-th band, and then perform the process of transmitting.

The acoustic decoding method of the present invention is an acoustic decoding method of receiving a packet storing a code by acoustic coding and decoding the code, wherein the acoustic coding is a scalable coding having a plurality of layers. The packet is set for each layer, the process of inspecting the presence or absence of packet loss in packet transmission, and for each layer in which no packet loss is detected, the code is extracted from the packet of the layer and decoded. The process has a process of obtaining a decoded signal, a process of adding each decoded signal as silence for a layer in which packet loss is detected, and a process of reproducing sound based on the result of the addition.

A first acoustic coding apparatus of the present invention is an acoustic coding apparatus that codes and transmits an acoustic signal, and is a means for performing scalable coding on an input acoustic code, and in the scalable coding. It has means for packetizing the code obtained for each layer and means for distributing the packet of each layer. Here, before transmission, it is preferable to further provide a means for performing error protection so that a packet of a layer of higher importance has a higher priority.

A second acoustic coding apparatus of the present invention is an acoustic coding apparatus which codes an acoustic signal for each of N (N is an integer of 2 or more) bands, and i is 2 or more and N or less. As each integer, the i-th band includes the (i-1) -th band, and a first extractor for extracting a component signal of the first band of the acoustic signal;
A first component to encode a component signal in the band
, And the i-th code based on the first to (i-1) th codes.
−1 th decoder for obtaining a decoded signal in the −1 band, an i th extractor for extracting a component signal in the i th band of the acoustic signal, a component signal in the i th band and the i th − The i-th code is obtained by encoding the residual signal with the decoded signal in the 1-band
Encoder and means for transmitting the first to Nth codes separately for each of the bands. An audio encoding device having.

An acoustic decoding apparatus of the present invention is an acoustic decoding apparatus that receives a packet storing a code by acoustic coding and decodes the code, and the acoustic coding is a scalable coding having a plurality of layers. Then, the packet is set for each layer, means for inspecting the presence or absence of packet loss in packet transmission, and for each layer in which no packet loss is detected, the code is extracted from the packet of the layer and decoded, It has means for obtaining a decoded signal, means for adding each decoded signal as silence for the layer in which packet loss is detected, and means for reproducing sound based on the result of the addition.

That is, according to the present invention, when packetizing a code bit string of a voice or a musical tone for distribution, the code information is layered by scalable coding, and the code bit string is independently packetized and distributed for each layer. By this method, even if a part of the delivery packet is lost, the sound is not interrupted, and the large distortion or noise hardly occurs. Furthermore, when sending a scalable code bit string in this way, error protection is focused on packets in a layer that contains major sounds that are easily heard by humans, or layers are prioritized to create a layer of high importance. By implementing strong error protection on the packet, it is possible to further suppress the sound quality deterioration due to the packet loss.

In the present invention, the packet loss includes that a packet does not reach the receiving side within a predetermined delay time.

[0016]

BEST MODE FOR CARRYING OUT THE INVENTION Next, preferred embodiments of the present invention will be described with reference to the drawings.

In the acoustic coding method of the present invention, scalable coding is adopted and packetization is performed for each layer in scalable coding. As a result, even if there is a packet loss, it is rare that multiple packets belonging to the same coded frame are lost at the same time, and the waveform in either band remains, resulting in significant sound quality. It is possible to prevent deterioration. Furthermore, if the packet corresponding to the band containing the most important acoustic component is transmitted with high priority for transmission error and packet loss, the transmission error and packet loss in the packet corresponding to the most important acoustic component will be transmitted. Is prevented, and even if a transmission error or packet loss occurs in the other packets, the sound quality is almost perceived as if it is not deteriorated.

An appropriate value can be adopted as the number of layers in the scalable coding. Here, if the number of layers in scalable coding is N (where N ≧
2) In the acoustic coding method according to the present invention, specifically, i is an integer of 2 ≦ i ≦ N, and the input acoustic signal includes the i-th band including the i−1-th band. A first extraction step of extracting the component signal of the first band of the acoustic signal as a division into N bands and a first extraction step of encoding the component signal of the first band to obtain the first code An encoding process, an (i-1) th decoding process for obtaining a decoded signal in the (i-1) th band based on the first code to the (i-1) th code, and a component signal in the i-th band of the acoustic signal are performed. The i-th extraction process of extracting, and the i-th
Performing the i-th encoding process of encoding the residual signal of the component signal of the band and the decoded signal of the i-1th band to obtain the i-th code, and if necessary, the i-1th For the N codes (the first code to the Nth code) obtained in this way so that the transmission error or omission in the band of is less than the transmission error or omission in the i-th band. Perform processing. When performing the i−1 th decoding process, if i> 2, if a decoded signal in the i−2 th band is required at this point, due to the hierarchical nature of the code in scalable encoding, A decoded signal in the i-1th band can be obtained by adding a decoded signal in the i-2th band to a signal obtained by decoding only the i-1th code. The compression technique used in the first encoding process, the compression technique used in the second encoding process, ..., And the compression technique used in the Nth encoding process may be different from each other. From the viewpoint of layered coding, an optimal coding method and compression method should be adopted according to each layer.

FIG. 1 is a diagram for explaining an acoustic coding method according to the present invention, and here, scalable coding is shown as an example in three layers. A layer that covers a main band of a sound is a main layer A, and its auxiliary layers are B and C. The main band of sound is, for example, a band of about 0 to 4 kHz in the case of a sound source mainly composed of voice, and is, for example, 0 to 0 in the case of a sound source mainly composed of music (or musical sound).
It is about 8 kHz. The sub-tier B completely contains the band of the main tier A and the sub-tier C completely contains the band of the sub-tier B (thus the sub-tier C also completely contains the band of the main tier A). Of course, the number of layers in scalable encoding may be any number as long as it is 2 or more.

A waveform is cut out from an input signal (acoustic signal) at a predetermined time interval (encoding frame unit) (step 1
01), perform scalable coding (step 10)
2). As scalable coding, the above-mentioned Patent No. 3
Treatments such as those disclosed in 139602 may be used. Here, it is assumed that the power spectrum of the input signal is as shown in FIG. 200 corresponding to the layers A to C. In this power spectrum, the area of the layer B, which is located in the elongated area of the layer A, corresponds to the coding residual after the coding of the layer A.

By the scalable coding, the waveform cut out from the original sound in coding frame units is coded, the code bit string A is generated from the main layer A, the code bit string B is generated from the auxiliary layer B, and the code bit string is generated from the auxiliary layer C. C occurs. These are scalable bit strings.

The code bit string thus obtained is then packetized (step 103). When packetizing this scalable code bit string, each layer is an independent packet. As a result, packets A to C are obtained corresponding to the code bit strings A to C. The packets A to C thus obtained are subjected to error protection as necessary as described later (step 104),
It is sent out and transmitted on the transmission line 110 (step 10).
5). Since each layer A to C has an independent packet, these packets A to C are also provided on the transmission line 110.
C will be transmitted independently for each layer. Such packets reach the receiving side via a line such as the Internet.

Although all the packets of the main layer A and the auxiliary layers B and C are transmitted here, the packets of all the layers are not transmitted, but from the main layer to a predetermined layer. It is also possible to transmit the packet of. The processing on the receiving side will be described later, but the receiving side does not perform decoding based on all the received packets, but according to the required sound quality,
It is possible to reproduce sound by decoding using packets from the main layer to a predetermined layer. Also, in the above description, packetization is performed in units of one encoded frame, but packetization may be performed in 2-3 encoded frames. Further, although the packets A to C are independently transmitted, a plurality of packets are used so that the code of the main layer A and the code of the auxiliary layer B are combined into one packet and the code of the auxiliary layer C is one packet. One packet may be generated from the code of the hierarchy. When one packet is generated from codes of a plurality of layers, if the codes of all layers are set to one packet, serious sound quality deterioration will occur when packet loss occurs. Although the number may be small, a plurality of packets are generated in the same time zone in the input acoustic signal.

During packet delivery, there are cases where packets are lost, such as congestion occurring on the network due to concentration of traffic, and failure in absorbing received packet jitter. When such a packet loss occurs, in a normal non-scalable coding method, one coded frame or a plurality of coded frames is packetized as one packet, and since these packets include information of the entire band, packet loss is reduced. In a certain coded frame portion, the entire band is lost and the sound quality is significantly deteriorated. However, in the case of this embodiment, it is considered rare that all of the packets A, B, and C constituting one encoded frame are lost at the same time, and therefore, as shown in FIGS. In addition, the waveform of some band remains without loss. Therefore, the sound quality hardly deteriorates significantly. Assuming that the acoustic signal whose power spectrum is shown in FIG. 1 is scalable encoded and transmitted, FIG. 2A shows the power spectrum when the packet of the main layer A is lost, and FIG. 2B shows the auxiliary layer. FIG. 2C shows a power spectrum when the packet of B is lost, and FIG. 2C shows a power spectrum when the packet of the auxiliary layer C is lost.

In the example described here, due to the characteristics of scalable coding, if the packet A corresponding to the main layer A remains without loss, the packets (packets B and C) of other layers will be lost. However, the sound quality will hardly deteriorate. Therefore, the error protection of step 104 in FIG. 1 is performed, and in the error protection, the packets of the important layers are prioritized so that the packets are less likely to be lost in the transmission path. It should be set to. In the example described here, the packet A that constitutes the main element of the acoustic signal has a high priority,
Next, the main packet B has a medium priority, and the most auxiliary packet C has a low priority. Then, based on such a priority, a high degree of error protection is performed for the packet A having a high priority so that the rate of packet loss on the receiving side becomes an extremely small predetermined value or less, The medium error protection is performed for the medium packet B of which the packet loss ratio on the receiving side is smaller than that of the case where the error protection is not performed, but larger than that of the packet A. For the packet C having a low priority, the light error protection is performed or the error protection is not performed. Alternatively, by using QoS (Quality of Service) technology or the like, there is a method in which the priority of the packet A is set higher so that the destination is reached earlier, and the packet C is set such that the packet C may be reached slowly. .

Various methods such as redundant code and retransmission are known as error protection methods for preventing packet loss on the receiving side when an error occurs on a transmission line.
In the present invention, an appropriate error protection method can be used. For example, if the "lost packet" is "a packet that could not reach the receiving side within a certain delay time (for example, 500 milliseconds)", the first layer, which is the most important layer (main layer A here) Is sent from the receiving side to the transmitting side to retransmit the packet, or double or triple packet transmission is performed. In addition, TCP (transmissio), which is a protocol incorporating error resending and error correction processing only in the first layer.
n control protocol), and the other layers (auxiliary layers B and C here) do not include error retransmission and error correction processing.
There is also a method of transmitting using a ram protocol). Furthermore, for example, as described in Japanese Patent No. 3212123 (or Japanese Patent Application Laid-Open No. 5-281998), a plurality of error correction encoders and error correction decoders having different correction capabilities and detection capabilities are used, For the packets of the first layer, error correction coding with high correction capability and detection capability, such as Reed-Solomon code, is performed, and for the second layer (auxiliary layer B here), the first layer is used. The correction ability and the detection ability are lower than those used in, for example, BCH (Bose-Ch
There is also a method in which error correction coding is performed using an audhuri-Hocquennghem) code or the like, and error correction coding is not performed for the third layer (auxiliary layer C here). A method that only packet A is transmitted via a high-speed and reliable leased line, and packets B and C are transmitted via the Internet, which causes some delay or error.

FIGS. 3A to 3C show the power spectrum obtained on the receiving side when such error protection is performed, assuming that the acoustic signal whose power spectrum is shown in FIG. 1 is scalable encoded and transmitted. Is shown. Figure 3 (a)
Shows a case where packet B and packet C are lost, and FIG.
3B shows the case where the packet C is lost, and FIG.
Indicates the case where packet B is lost. Since the packet A for the main band is protected so as not to be lost as much as possible, the waveform remains in the main band even if the packet loss occurs, and therefore the sound quality is hardly deteriorated.

Next, such error protection will be described. Here, a configuration will be described in which, when the code bit string having a scalable hierarchical structure is packetized for each layer and distributed by the three-layer scalable coding, the sound quality is not deteriorated much even if a packet loss occurs. To do. As shown in FIG. 4A, a transmission side (encoding device) 11 and a reception side (decoding device) 12 are connected to the network 10.
Are connected.

As shown in FIG. 4B, first, before distributing the acoustic signal code packet, the congestion degree of the network 10 is investigated and predicted by some method between the transmitting side or the receiving side or between the transmitting and receiving (step 131). ), And obtain a packet loss estimate (step 132). The congestion degree of the network is a parameter indicating how much communication is occupied in the network. For example, when the TCP protocol is used, packets are transmitted between the transmission side and the reception side. It can be predicted by monitoring or by observing how much a test packet is delivered before the actual transmission and is dropped at the receiving side. When the receiving side investigates the congestion level of the network, the receiving side notifies the transmitting side of the investigation result. Then, according to the estimated packet loss rate, error protection is set for each layer of the scalable coding so that the reception success probability to the receiving side is guaranteed to be a specified value or more (step 133), and each layer is set. The packet is delivered after the error protection information is given based on such a set value (step 134).

To give an example with specific numerical values, here, for simplicity, assuming that the packet size of all packets (before error protection processing) is the same, when the average packet loss rate estimated value is 15% or less, In the probability of successful reception of the packet to the receiving side, the packet of the main layer A is guaranteed to be 99% or more, the packet of the auxiliary layer B is 95% or more, and the packet of the auxiliary layer C is 91% or more, or , If the average packet loss rate estimated value is 16% or more and 55% or less, in the reception success probability of the packet to the receiving side, the packet of the main layer A is 95% or more, the packet of the auxiliary layer B is 80% or more, and the auxiliary Designation is made such that the packet of the layer C is guaranteed at 70% or more.

Then, it is judged whether or not the transmission is completed (step 135). If the transmission is completed, the process is terminated. If the transmission is not completed, error protection information is given to the next packet and the packet is transmitted. Return to step 134.

By performing such an error protection process, the mechanism of preferentially protecting the sound of the main layer can be achieved. Therefore, when decoding at the receiving side, a packet having a packet loss rate of the above estimated value is used. Even if a loss occurs, the degree of sound quality deterioration can be reduced.

When prioritizing the reception success rate of packets, a communication protocol that guarantees a reception success probability that can be approximated to 100% within a certain delay condition on the receiving side by the error protection information. It is assumed to be used. In that case, a packet given such error protection information can be regarded as one that can be successfully received. Even in that case, the average packet loss rate estimation value is obtained before packet delivery through the same process as the above-mentioned example of the error protection process, and if packet loss is expected to occur, the packet of the main layer A For error packet, the error protection information is given so that the reception success probability becomes a value that can be approximated to 100%.
The packet loss state is controlled so that the packet loss occurs in C and C. For example, when the estimated average packet loss rate is 15% or less, the packets of the main layer A are guaranteed not to be substantially lost, and the packets of the auxiliary layer B of 95% or more and the auxiliary layer C of 90% or more are guaranteed. The error protection information is given to the delivery packet so that the delivery success probability of is obtained. Alternatively, when the estimated average packet loss rate is 16% or more and 55% or less, the packet of the main layer A is guaranteed not to be lost and the auxiliary layer B is
Is 85% or more and the auxiliary layer C is 70% or more, the error protection information is given to the distribution packet and the packet is distributed.

By performing such error protection, the packet of the main layer A is surely protected, and even if a packet loss occurs on the transmission line, the packet shown in FIG.
As shown in (c) to (c), the sound quality of the main part (band) is free from deterioration, so that good quality is maintained.

Next, an acoustic coding apparatus for performing the above acoustic coding will be described. Specifically, such an audio encoding device, for example, when the number of layers in scalable encoding is N (where N ≧ 2), i is 2 ≦ i
Each integer of ≦ N is set, and the input acoustic signal is divided into N bands so that the i-th band includes the i−1-th band, and the component signal of the first band of the acoustic signal is extracted. First
Of the first band, the first coder for coding the component signal of the first band to obtain the first code, and the first code to the (i-1) th code
Of the i-th band of the i-th band of the acoustic signal, the i-th decoder for obtaining the decoded signal of the i-th band based on the code of Component signal and i-th
-I encoder that encodes the residual signal with the decoded signal of the -1 band to obtain the i-th code, and packetization that packetizes the codes (first to N-th code) of each layer And a multiplexing unit for sending these packets to the transmission path. A total of N-1 number of i-1th decoders, a total of N-1 number of i-th extractors, a total of N-1 number of i-th encoders, and a total number of packetizers are provided. N pieces are provided. Further, if the above-mentioned error protection is performed, an error protection unit (total N) for performing error protection processing on each packet is provided corresponding to each layer,
Each error protection unit performs error protection processing so that transmission errors or omissions in the i-1th band are less than transmission errors or omissions in the i-th band.

FIG. 5 shows a concrete configuration of an acoustic coding apparatus used when N = 3, that is, when scalable coding of three layers in the above example is performed. Here, the first band (the band of the main layer A) is set corresponding to the main band of the sound, and the second band (of the auxiliary layer B) wider than the first band is included so as to include the first band. It is assumed that a third band (a band of the auxiliary layer C) wider than the second band is set so as to include the second band.

This acoustic coding apparatus includes a clipping unit 21 for clipping a waveform from an input signal (acoustic signal) to obtain a coded frame, and a first band (the band of the main layer A) for extracting a component signal. 1 extractor 22, a first encoder 23 that obtains a first code by encoding the component signal of the first band, and a first band that decodes the first code to decode the first band. A first decoder 24 for obtaining a signal, a second extractor 25 for extracting a component signal of a second band (band of auxiliary layer B),
The subtractor 26 that generates a residual signal by subtracting the decoded signal of the first band from the component signal of the second band, and the residual signal obtained by the subtractor 26 is encoded and A second encoder 27 for obtaining a code, and a second encoder 27
Second decoder 28 for decoding the second code output from
And an adder 29 for adding the above-mentioned decoded signal of the first band and the output signal of the second decoder 28 to obtain a decoded signal of the second band, and a third band (band of auxiliary layer C) The third extractor 30 for extracting the component signal of the second band, the subtracter 31 for generating the residual signal by subtracting the decoded signal of the second band from the component signal of the third band, and the subtracter 31 A second encoder 32 that encodes the residual signal to obtain a third code, a first packetizer 33 that packetizes the first code, and a packetize the second code. A second packetizer 34, a third packetizer 35 that packetizes a third code, and a first error protection that performs error protection processing on the output packet of the first packetizer 33. Part 36
A second error protection unit 37 that performs error protection processing on the output packet of the second packetizer 34, and a third error protection unit 37 that performs error protection processing on the output packet of the third packetizer 35. An error protection unit 38 and an error protection control unit 39 that sets the level of error protection processing in each of the error protection units 36 to 38.
And a multiplexing unit 40 that multiplexes the packets of each layer that have been subjected to error protection processing and sends them out onto the transmission path.
In each of the error protection units 36 to 38, the error protection with the highest priority is given to the packet of the first band, and the error protection of the next priority is given to the packet of the second band. The error protection control unit 39 sets the level of the error protection processing so that the packet of the third band is subjected to the error protection with the lowest priority (including the case where the error protection is not executed).

Next, the decoding process on the receiving side will be described assuming that acoustic coding has been performed on the transmitting side as described above.

When the receiving side receives a packet from the transmission line, it first sorts the packet into each layer. And
The packet is inspected for each layer, and if the error protection information is added, error detection and error correction are performed to obtain the code of the layer. Then, the output acoustic signal is obtained by decoding the code of each layer, adding each decoded signal and reproducing the acoustic signal. At this time, if there is a packet loss (a packet that has not arrived at the receiving side by the time limit or a packet that has an error that cannot be corrected by error correction), only the layer with packet loss is treated as having no packet. Therefore, even when each decoded signal is added, the hierarchy is not set as an addition target. That is, it is assumed that the layer is silent. Even with this processing, as described with reference to FIG. 2 or FIG. 3, since the acoustic signal is reproduced from the waveforms of the remaining layers, significant deterioration in sound quality of the output acoustic signal can be avoided. In particular, if sufficient error protection is provided for the packets of the main layer, the packets of the main layer are practically not lost, and therefore the deterioration of the sound quality due to the packet loss is significantly reduced. It here,
I explained that the layer with packet loss is silenced,
If there is a packet loss, the packet of the coded frame immediately before that layer may be used again,
The arithmetic mean of the packet of the immediately preceding encoded frame and the packet of the immediately following encoded frame in the layer may be obtained and the signal sequence of the arithmetic average may be used.

In such an acoustic decoding device, assuming that the number of layers in scalable encoding is N and the acoustic signal is divided into N bands as described above (where N ≧ 2), j If each integer of 1 ≦ j ≦ N is satisfied, a j-th code check unit that detects a code error or a loss in the j-th code and a j-th code that is decoded when neither an error nor a loss is detected. To obtain the decoded signal of the j-th band by the following: an adder that adds the decoded signal of the j-th band in which no error or omission was detected over the band; and an addition component obtained by the adder. And an acoustic signal reproducing unit that reproduces an acoustic signal based on the sound signal.

Assuming that N = 3, such an audio decoding apparatus, as shown in FIG. 6, has a code separator 51 for separating a packet in a received signal into layers and a packet for the first layer. A first code checking unit 52 for checking for errors and omissions and extracting a first code, and a first code checking unit 52
First if no error or omission was detected in
A first decoder 53 that decodes the code of 1 to obtain a decoded signal of the first band, and a second code checking unit 54 that checks the packets of the second layer for errors and omissions and extracts the second code. A second decoder 55 that decodes the second code to obtain a decoded signal in the second band when no error or omission is detected in the second code checking unit 54; Third code inspecting unit 56 for inspecting the third packet for errors and omissions, and decoding the third code when no error or omission is detected in the third code inspecting unit 56. A first decoder 57 for obtaining a decoded signal of the third band by means of the above, an adder 58 for adding the decoded signals of the first to third bands over the band, and an acoustic signal based on the addition result of the adder 58. And an acoustic signal reproducing unit 59 that performs a reproducing process. That. However, in the adder 58, only the decoded sound of the layer A is used, the decoded sound obtained by adding the layers A and B is used, or all of the layers A, B, and C are used in the adder 58. The added decoded sound can be used.

Although the preferred embodiment of the present invention has been described above, the above-described acoustic encoding device and acoustic decoding device may have a hardware configuration, or a general-purpose computer or microprocessor may be used. It can also be configured by software.

That is, both the acoustic coding apparatus and the acoustic decoding apparatus load a program for executing the acoustic coding method and the acoustic decoding method described above into a computer such as a one-board computer including a microprocessor and the like. , Can also be realized by executing the program. A program for performing the acoustic encoding method and the acoustic decoding method is read into a computer by a recording medium such as a CD-ROM or a non-volatile memory, or via a network. This computer reads, for example, a CPU such as a microprocessor, a memory for storing programs and data, an input / output device for inputting and outputting acoustic signals, a network interface for connecting to a network, and a recording medium. And a reading device. Memory, input / output device,
Both the network interface and the reading device are connected to the CPU. In this computer, a program for performing acoustic coding or acoustic decoding is read from a recording medium or a network, loaded on a memory, and the CPU executes the program to perform the acoustic coding or acoustic decoding as described above. Is executed.

[0044]

As described above, according to the present invention, since the acoustic signal is encoded by the scalable coding and the packet corresponding to each layer in the scalable coding is distributed in each layer. There is an effect that it is possible to prevent a significant deterioration in sound quality when a packet loss occurs. Further, by sufficiently protecting the packets of the important layer from error, even if a packet loss occurs, the deterioration of the sound quality can be minimized. Therefore, according to the present invention, when packetized musical sound or voice data is distributed through various lines such as the Internet, it can be expected that transmission without feeling sound breakage or deterioration is performed. In particular, in a communication system that requires real-time transmission, such as live programs in Internet radio broadcasting and Internet TV broadcasting, Internet telephones, and Internet conferencing systems, retransmission cannot be performed when packet loss occurs.
According to the present invention, it is possible to expect realization of live communication in which quality is not deteriorated or almost no deterioration is felt even if a lost packet is not retransmitted.

[Brief description of drawings]

FIG. 1 is a diagram illustrating an acoustic coding method according to an embodiment of the present invention.

FIGS. 2A to 2C are diagrams showing a reproduction sound spectrum when a packet in each layer is lost.

3 (a) to 3 (c) are diagrams showing a reproduction sound spectrum when a packet corresponding to a main layer A is subjected to a high degree of error protection and a packet in another layer is lost. .

4A is a block diagram showing a network configuration, and FIG. 4B is a flowchart showing a procedure for setting error protection information.

FIG. 5 is a block diagram illustrating an audio encoding device according to an embodiment of the present invention.

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

[Explanation of symbols]

10 network 11 sender 12 Receiver 21 Cutout part 22, 25, 30 extractor 23, 27, 32 encoder 24, 28, 53, 55, 57 Decoder 26,31 Subtractor 29,58 adder 33-35 Packetizer 36-38 Error protection unit 39 Error protection control unit 40 Multiplexer 51 code separator 52, 54, 56 Code checking unit 59 Acoustic signal playback unit 110 transmission line

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Kazunaga Ikeda             2-3-1, Otemachi, Chiyoda-ku, Tokyo             Inside Telegraph and Telephone Corporation (72) Inventor Takeshi Mori             2-3-1, Otemachi, Chiyoda-ku, Tokyo             Inside Telegraph and Telephone Corporation F-term (reference) 5D045 DA01 DA11                 5K028 AA12 EE08 KK32 MM09 SS05                       SS15

Claims (13)

[Claims] 1. An acoustic encoding method for encoding and transmitting an acoustic signal, wherein scalable coding is performed on an input acoustic code, and the code obtained for each layer in the scalable encoding is packetized. An audio encoding method for delivering packets of each layer. 2. The acoustic coding method according to claim 1, wherein the packet of each layer is distributed after performing error protection processing so that a packet of a layer having a higher importance has a higher priority. 3. An acoustic encoding method for encoding an acoustic signal for every N (N is an integer of 2 or more) bands, wherein i is an integer of 2 or more and N or less, and the i-th band is i-
A first extraction step of extracting a component signal of the first band of the acoustic signal including a first band; and a first encoding for encoding the component signal of the first band to obtain a first code. An i-1 th decoding step of obtaining a decoded signal of the i-1 th band based on the first to i-1 th codes, and extracting a component signal of the i th band of the acoustic signal An i-th code for obtaining an i-th code by encoding an i-th band component signal and a residual signal of the i-th band component signal and the i-1 th band decoded signal.
And the step of transmitting the first to Nth codes separately for each of the bands and transmitting the acoustic coding method. 4. An error protection process is performed so that transmission errors or omissions in the (i-1) th band are less than transmission errors or omissions in the i-th band, and then the step of transmitting is performed. The audio encoding method according to item 3. 5. An acoustic decoding method for receiving a packet storing a code by acoustic coding and decoding the code, wherein the acoustic coding is scalable coding having a plurality of layers, and the packet Is set for each layer, and the process of inspecting for packet loss in packet transmission and for each layer in which no packet loss is detected, the code is extracted from the packet of the layer and decoded to obtain a decoded signal. As for the process and the layer where packet loss is detected, as silence,
An acoustic decoding method comprising: a step of adding the respective decoded signals; and a step of reproducing a sound based on a result of the addition. 6. An acoustic decoding method for receiving a packet storing a code by acoustic coding and decoding the code, wherein the acoustic coding is scalable coding having a plurality of layers, and the packet Is set for each layer, and the process of inspecting for packet loss in packet transmission and for each layer in which no packet loss is detected, the code is extracted from the packet of the layer and decoded to obtain a decoded signal. And a step of adding a decoded signal using a decoding result of a packet in at least one of the coded frames immediately before and after the same layer with respect to a layer in which packet loss is detected, and a step of adding the respective decoded signals. And a step of reproducing sound based on a result of the addition, the sound decoding method. 7. A j-th code check process for detecting a code error or loss in the j-th code, where N is an integer of 2 or more and j is an integer from 1 to N, and the j-th code If no error or omission is detected, the j-th decoding process for decoding the j-th code to obtain a decoded signal in the j-th band, and the decoded component of each band in which neither error nor omission is detected And an acoustic signal reproduction step of reproducing an acoustic signal based on the addition component obtained in the addition step, wherein the j-th code is the j-th band in scalable encoding. And an i-th band includes an i-1-th band, where i is an integer from 2 to N. 8. An acoustic encoding device for encoding and transmitting an acoustic signal, comprising means for performing scalable encoding on an input acoustic code, and a code obtained for each layer in the scalable encoding. An audio encoding device comprising: a packetizing unit; and a unit for delivering the packet of each layer. 9. The audio encoding device according to claim 8, further comprising means for performing error protection so that a packet of a layer having a higher importance has a higher priority before transmission. 10. An acoustic encoding device for encoding an acoustic signal for every N (N is an integer of 2 or more) bands, wherein i is an integer of 2 or more and N or less, and the i-th band is i-
A first extractor that includes a first band and that extracts a component signal of a first band of the acoustic signal; and a first encoding that encodes the component signal of the first band to obtain a first code. An (i-1) th decoder that obtains a decoded signal in the (i-1) th band based on the 1st to (i-1) th codes, and a first signal that extracts a component signal in the i-th band of the acoustic signal. an i-th extractor, and an i-th code for obtaining a i-th code by encoding a residual signal between the i-th band component signal and the i-1 th band decoded signal.
And an encoding unit that transmits the first to Nth codes separately for each of the bands. 11. An acoustic decoding device for receiving a packet storing a code by acoustic coding and decoding the code, wherein the acoustic coding is scalable coding having a plurality of layers, and the packet Is set for each layer, and means for inspecting for packet loss in packet transmission and, for each layer in which no packet loss is detected, extracts a code from the packet of the layer and decodes it to obtain a decoded signal. As a means and silence for the layer where packet loss is detected,
An audio decoding apparatus comprising: a unit that adds each of the decoded signals; and a unit that reproduces a sound based on the result of the addition. 12. A program for causing a computer to perform acoustic encoding for encoding and transmitting an acoustic signal, the computer performing scalable encoding on an acoustic code to be input, and the scalable encoding. Packetizing the code obtained for each layer in, the process of performing error protection so that the packet of the layer of higher importance has higher priority, and the process of delivering the packet of each layer. Program to be executed. 13. A program for causing a computer to perform acoustic decoding for receiving a packet storing a code by acoustic coding and decoding the code, wherein the acoustic coding is scalable coding having a plurality of layers. The packet is set for each layer, and a process for inspecting the computer for the presence or absence of packet loss in packet transmission, and for each layer in which no packet loss is detected, the packet is encoded from the packet of the layer. Is extracted and decoded to obtain a decoded signal, and regarding the layer where packet loss is detected, there is no sound,
A program for executing a process of adding the respective decoded signals, and a process of reproducing sound based on a result of the addition.