JPH0414334A - Packet interpolation system - Google Patents

Abstract

PURPOSE:To process sub band ADPCM codes for plural bands into packets efficiently, to transmit and decode them in an excellent way by generating a multiplex packet from ADPCM codes for plural bands in one and the same time, decomposing them at a receiver side and applying packet interpolation to the decomposed signal only in the case of aborting them. CONSTITUTION:A sub band ADPCM coder 53 of a packet transmission terminal equipment 51 divides an input voice signal or the like into plural frequency bands for a same period to form the signal into a relevant sub ADPCM codes and a packet composition section 54 composes the sub ADPCM codes into a multiplexed packet being one kind of a packet signal, which is sent via a packet network 50. Then a packet decomposing section 56 of a packet reception terminal equipment 52 decomposes the packet into a sub ADPCM code packet and an interpolation device 57 applies packet interpolation to the code in the case of the presence of packet abort and a decoded signal is outputted via a decoder 58. Through the multiplexer constitution above, the sub band ADPCM codes for plural bands are processed into packets efficiently, they are sent, received and decoded in an excellent way.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a packet interpolation method in packet communications, particularly in voice band packet transmission where band compression is performed by high-efficiency encoding such as ADPCM encoding. This invention relates to a packet interpolation method for interpolating packets.

(Prior art) For example, the CCITT (International Telegraph and Telephone Consultative Committee) recommendation regarding ADPCM encoding currently specifies a high-efficiency encoding method for transmitting signals in the 3 and 4 kHz band, which is the telephone band, at 32 kbit/S. Recommendation G, 721 and 7.0
There is Recommendation 0.722 which specifies the subband (band division) ADPCM encoding method for transmitting a kHz wideband signal at 64 kbit/s. Furthermore, application of subband ADPCM encoding to voice band signals in the 20 kHz band is being considered.

On the other hand, with the development of digital communication networks, packet communication systems that handle voice signals, audio band signals, image signals, data signals, etc. in an integrated manner are being considered. In voice communication using the packet communication method, improvement in transmission efficiency can be expected by transmitting only the sound sections of voice signals generated in eight bursts. Furthermore, it is possible to flexibly set the transmission rate according to desired quality.

However, in packet communications, packets may be discarded due to traffic congestion within the network, and when this occurs, the quality of the reproduced signal deteriorates on the receiving side. In particular, in an example where a high-efficiency encoding method such as the aforementioned PC:M encoding is applied to signal encoding, a mismatch in ADPCM code I (k) may occur between the encoder and decoder due to packet discard. , an asynchronous state, which is a mismatch between the internal states of the adaptive quantizer and the adaptive predictor, may occur, resulting in significant quality degradation. Therefore, a parker that suppresses this asynchronous state caused by packet discards and compensates for quality deterioration.

An interpolation method is required.

Conventionally, as a packet interpolation method when full-pant ADPCM encoding is applied to packet communication, a method that focuses on the periodicity of a residual signal such as an audio signal has been proposed. There is an application entitled "Packet Interpolation Method" filed in Japanese Patent Laid-Open No. 1-18E11. The proposed method is a method in which interpolation processing is performed on the ADPCM code received at the packet receiving terminal side.

In a packet communication system using ADPCM encoding, an input signal is encoded into an ADPCM code at a packet transmitting terminal side, and the ADPCM code is packetized by a packet assembling unit and transmitted to a packet network. On the packet receiving terminal side, first, the packet transmitted from the server network is
Decompose into DPCM code, then convert this ADPCM code into AD
Input to PCH1 code interpolator. The ADPCM code interpolator interpolates the missing part due to packet discard.
Output PCM code.

This interpolated ADPCM code is input to an ADPCM decoder, and a reproduced signal is output from the ADPCM decoder.

(Problem to be Solved by the Invention) In the full punt interpolation method described above, the entire band is
It is encoded and transmitted as a packet. The receiving side restores this and performs interpolation if any packets are discarded. However, in subhand ADPCM encoding, since there are multiple bands, the transmitting side packetizes the multiple bands, and the receiving side restores the multiple bands. That is, since packetization on the transmitting side and restoration and interpolation on the receiving side cover multiple bands, there is a problem that the interpolation method for one band in the full/hand system cannot be applied as is to the subband system.

The present invention solves the above problems and
It is an object of the present invention to provide a packet interpolation method that can efficiently packetize and restore multiple bands of M codes.

(Means for Solving the Problems) According to the present invention, packets that interpolate packet discards in a packet communication system that transfers packets containing voice band signals encoded in ADPCM codes from a transmitting side to a receiving side through a packet network. In the interpolation method, the transmitting side divides the audio band signal into multiple frequency bands within the same period, encodes them into corresponding DPCM codes of multiple bands, and stores the ADPCM codes of multiple bands within the same period in the information section. Then, a header is added to the information part and assembled into a multiplexed packet, and the multiplexed packet is sent to the receiving side through the packet network. When the receiving side receives the multiplexed packet, it converts the multiplexed packet into ADPCM codes of multiple bands. This decomposed A
If DPCM codes are temporarily stored and no packets are discarded,
selecting the decomposed ADPCM codes of multiple bands, and when a packet is discarded, selecting the stored ADPCM codes for a predetermined period, decoding and combining the ADPCM codes of the multiple bands thus selected; Outputs audio band signal.

According to the present invention, the packet transmitter in the packet communication system as described above divides the voice band signal into a plurality of frequency bands within the same period and performs ADPC of the plurality of bands.
A sub-end ADPCM encoding means for encoding into an M code and ADPCM codes of multiple bands within the same period are stored in an information section, a header is added to the information section, assembled into a multiplexed packet, and transmitted to a packet network. and packet assembly means.

Further, according to the present invention, the packet receiving device in the above-mentioned Path-2 communication system includes a packet assembling means that decomposes the multiplexed packet into ADPCX codes of multiple bands when receiving the multiplexed packet, and
An ADPCM that temporarily stores M codes, and if there is no packet discard, selects an ADPCM code of a plurality of bands output from the disassembly means, and if a packet is discarded, selects the stored ADPCM code for a predetermined period. code interpolation means, and a sub-panel for decoding and synthesizing the selected ADPCM codes of multiple bands and outputting a voice band signal.
``ADPCM decoding means.

(Function) According to the packet interpolation method according to the present invention, the transmitting means of the packet terminal divides the audio band signal into multiple bands, and
Encode into PCM code. Then, the ADPCM codes of multiple bands within a predetermined period are assembled into multiplexed packets and output to the packet network.

When the receiving means of the packet terminal receives this multiplexed packet, it decomposes it into ADPCM codes of multiple bands and inputs them to an ADPCM code interpolator corresponding to these multiple bands. The low-pass interpolator accumulates the input ADPCM code for a predetermined period and estimates the pitch period of the code. The low-band interpolator normally outputs the input ADPCM code, but when a packet is discarded, it compensates for the loss of the packet by outputting the ADPCM code with the estimated pitch period corresponding to the discard period as an interpolation code. do. Other high-frequency interpolators are
An estimate representing the estimated pitch from the low-pass interpolator is input;
According to this, high frequency interpolation is performed in the same way as low frequency.

(Embodiment) Next, an embodiment of the packet interpolation method according to the present invention will be described in detail with reference to the attached drawings. According to the packet interpolation method according to the present invention, the transmitting side uses ADPCM codes of multiple bands as an information part. generate multiplexed packets. The receiving side disassembles this packet, and if the packet is discarded, the low-band AD
A pitch period is estimated from PCM encoding, and defects in ADPGM codes in other bands are interpolated using this pitch period.

Conventionally, as a packet interpolation method when full-pant ADPCM encoding is applied to packet communication, a method has been proposed that focuses on the periodicity of a residual signal such as an audio signal. First, the conventional packet interpolation method will be explained, and then the interpolation method according to the present invention will be explained.

FIG. 7 shows a configuration example of a packet communication system to which the packet interpolation method using PCM encoding according to the present invention can be applied. The packet transmitting terminal 8o encodes the input signal into an ADPCH code using a full/hand ADPCM encoder 81. This ADPCM code is packetized by the second packet assembling unit and transmitted to the packet network 85. In the packet receiving terminal 86 , the packet transmitted from the packet network 85 is decomposed into ADPCM codes by the packet assembling unit 87 , and the codes are input to the ADPCM code interpolator 88 . The ADPCM code interpolator 88 interpolates any missing parts due to packet discard, and
Outputs DPCM code. This interpolated code is ADP
The signal is input to the GN decoder 89, and the ADPCM decoder 88 decodes it and outputs a reproduced signal.

FIG. 8A shows a full-pant ADPC in this prior art.
An example of the configuration of the M encoder 81 is shown. When the code k is a discrete time, when the input signal s (k) and the predicted signal s (k) whose sign is inverted are input to the adder 61,
Both residual signals e (k) are output. Adaptive quantizer (
When e (k) is input, AQ) e2 performs ADPCM encoding on it and outputs ADPCM code I (k). The adaptive inverse quantizer (AQ”) 83 is the code I
When (k) is input, a quantized residual signal e (k) is output. The adder 64 receives this signal e (k) and the predicted signal s (k), and outputs a reproduced signal s (k). The adaptive predictor 65 outputs a predicted signal 5(k) when the reproduced signal s(k) is input.6 FIG. 8B shows an example of the configuration of a conventional full-length ADPCM decoder 89. The adaptive inverse quantizer (AQ) 8B is an ADPC
M code I (k) is input, and quantized residual signal e (k
) is output. When the adder 67 receives this signal e (k) and the predicted signal s (k), it outputs a reproduced signal s (k). The adaptive predictor (AP) 28 reproduces the reproduced signal s
(k) is input and a predicted signal s (k) is output.

Figure 9 shows ADP using the full pant ADPCM encoding method.
This is an example of the configuration of the CM code interpolator 88. The ADPCM code switching unit 71 monitors the state of packet discard occurrence using the packet discard occurrence instruction signal L(k) from the input 502. The switching unit 71 converts the input signal 1r(k) as it is when no packet discard has occurred, and converts the ADPCM code generation signal 1p(k) into the interpolated ADPCM code Ire(k) when the packet discard has occurred. ) is output to the output 503.

The buffer memory update unit 72 uses this interpolated ADPC
When the M code 1re(k) is input, it is updated and the updated buffer memory output data (IBUFF) is output. The pitch period estimation unit 73 uses this output data IB
A UFF is input and a pitch period estimate (PIT) is output. The pack-a-memory pointer specifying unit 74 receives this estimated value PIT and the above-mentioned updated pack-a-memory output data IBUFF, and thereby outputs the above-mentioned ADPCM code generation signal Ip(k) to the switching unit 71. The pack memory output data I BUFF stores decoded values of past interpolated ADPCM codes 1re(k), but the buffer memory update unit 72 sequentially stores the latest Ire(k) values.
(k) is input, and the output data IBO is determined by the following equation (1).
FF is updated.

IBUFF (m) = IBUFF (m-1) 2≦
m≦M (1)IBUFF(1)=Q-'(I
re(k)) where M is the size of the buffer memory and Q
'() indicates decoding processing.

The pitch period estimator 73 performs autocorrelation calculation processing and the like on the value of the output data IBUFF to obtain a pitch period estimated value PIT. - Examples include: Autocorrelation function R(
t) as shown in the following equation (2).

Here, T MIN and T WAX are the minimum number of samples and the maximum number of samples, respectively, for which a pitch period is estimated to exist. For example, TMIN = 24, TMA
Set X = 128 t, and N = 128 d.

In this case, the size of the buffer memory IBUFF is 129
becomes. The value of t at which this autocorrelation number R(t) becomes the maximum value is the pitch period estimated value PIT.

The packed memory pointer designation unit 74 uses this PIT to obtain the ADPCM code generation signal 1p(k) using equation (3).

Ip(k) = Q(IBUFF(PIT)) where Q(
) indicates encoding processing.

In this way, the ADPCM code interpolator 88 performs AI) PCM
The pitch period of the signal is extracted from the past ADPCM code by using the periodicity of the code, and in the packet discard section, the A of the previous pitch period stored in the buffer memory is used.
The DPCM code is output as an interpolated value. This makes it possible to minimize the degree of asynchronous state between the ADPCM encoder 81 and the decoder 89 at the time of packet discard, and to suppress deterioration in the quality of the reproduced signal.

The packet interpolation method based on the full-pan I'' ADPCM encoding method described above does not require priority control of nodes within the packet network. It also has advantages such as no delay time caused by interpolation.

Now, FIG. 1 shows an embodiment of a packet communication system using sub-end ADPCM encoding to which the packet interpolation method according to the present invention is applied. The packet communication system of this embodiment includes a packet transmitting terminal 51, which includes a subband ADPCM encoder 53 and a node assembling unit 54. The encoder 53 receives an input signal of a voice band signal through its input signal 53-1, divides it into a plurality of (n) bands, and sends the input signal to a plurality of ADPCM encoders tl(k) to Itn(k).
When this ADPCM code is input, the 0 assembling unit 54 outputs the ADPC of multiple bands occurring within the same time.
The M codes I tl(k) to I tn(k) are multiplexed and assembled into a single multiplexed packet pt, which is sent to the transmission path 5.
Send to 4-1. The packet Pt is the packet network 50
and arrives at the packet receiving terminal 52 via the transmission path 50-1.

The packet receiving terminal 52 has a packet assembling section 56 that accommodates a transmission path 50-1, and when the multiplexed packet pt is input from the packet network 50, the packet assembling section 56 assembles a plurality of ADPCM codes I rl ( k) ~Irn
Decompose into (k). The output of the same decomposition unit 56 is ADPCM.
A sign interpolator 57 is connected. When these ADPCM codes are input, the interpolator 57 interpolates any packet discards that occur in them and converts them into interpolated ADPC codes.
Output M codes I rcl(k) to I rcn(k).

The interpolated output of the interpolator 57 is sent to the subband ADPCM decoder 5.
8, and the latter is input to this ADPCM code I rcl (
k) to Ircn(k) and outputs the output signal of the voice band signal to its output 58-1.

Note that this specific configuration of subband ADPCM encoder 53 and decoder 58 is an example for explaining the present invention, and the present invention is not limited to this specific configuration. For example, the number of bits of the ADPCM codes Itl(k) to Itn(k) and the value of the number of band divisions n are determined by requirements such as the bandwidth of the input signal and the required quality, and are not limited by the present invention. . - To give an example, in the case of CCITT Recommendation G, 722 for the 7 kHz radio band mentioned above:
The number of divisions n is 2, and the number of bits per sample of the ADPCM codes Itl(k) and It2(k) are 6 bits and 2 bits, respectively.

FIG. 4A shows a configuration example of the subband I'' ADPCM encoder 53 in this embodiment, and FIG. 4B shows a configuration example of the subband ADPCM encoder 53 in this embodiment.
An example of the configuration of the CM decoder 58 is shown. Sub/Hend ADP
The CM encoder 53 uses QMF (Quadrature M
31, which receives an input signal s(k) from an input terminal 53-1, divides it into two bands by a frequency division filter, and outputs a low-frequency input signal 5l(k) and a low-frequency input signal 5l(k). High frequency input signal s2(k)
Output. The former signal 5l(k) is input to the low-band ADPCM encoder 32, which converts it into ADPCM
It encodes and outputs a low-band ADPCM code 11(k).

Further, the high-frequency input signal 52(k) is input to the high-frequency ADPCM encoder 33, which converts it into high-frequency ADPCM
It is ADPCM encoded into code I2(k) and output.

Referring to FIG. 4B, subband ADPCM decoder 5
8 has a low-band ADPCM decoder 34 and a high-band ADPCM decoder 35. The low-band ADPCM decoder 34 uses the low-band ADP interpolated by the ADPCM code interpolator 57 described above.
A CM code 11c(k) is input, which is reproduced into a low frequency reproduction signal 5l(k) and output. The high-frequency DPCM decoder 35 receives the high-frequency DPCM code I 2 (k), and reproduces and outputs a high-frequency reproduced signal 52 (k). The outputs of the low band ADPCM decoder 34 and the high band ADPCM decoder 35 are connected to QMF 3fl, which is
It receives the low frequency reproduction signal 5l(k) and the high frequency reproduction signal 52(k) and outputs them as the reproduction signal s(k).

In this embodiment, the ADPCM encoder 53 differs from the full-pan FADPCM R encoder 81 according to the conventional method described above.
It has DPCM encoders 32 and 33. Similarly, the subband ADPCM decoder 58 also has ADPCM decoders 34 and 35 for the low band and high band, respectively. In addition, the number of band divisions is 2 in this example, but QMF
It goes without saying that the structure of 31 and 36 allows division into the necessary number.

FIG. 3 shows an example of the configuration of the packet assembling section 54 of the transmitting terminal 51. As shown in FIG. The packet assembling unit 54 has /header memories Ml-Mn corresponding to the number of bands n, and has subband A
ADPCM code Itl(k) ~ from the DPCM encoder 53
It receives I tn(k) and temporarily MaIs them into the corresponding buffer memories old to Mn. Buffer memory old~Mn
The output of is connected to the multiplexing section 33. Multiplexing section 33
is the ADPCM code I tl(k
) ~I tn(k) to the information section 302 (Figure 5A)
The header generation section 32 also adds a header 300 to this to generate a multiplexed packet pt.

The header generation unit 32 generates such a header 300, and its reading is controlled by the readout control unit 31. In addition, ADPCM codes I of multiple bands occurring within the same time
tt(k) ~I tn(k) buffer 7 memory M
Reading from 1 to Mn is also controlled by the read control unit 31.

The header 300 of the packet pt includes various control information necessary for packet transfer within the packet network 50.

- For example, if the information field length is 48 bytes and the above-mentioned Recommendation G, 722 is applied, n.2. T=
48, and the information generated in 6 milliseconds is 1 packet p.
Included in t. FIG. 5A shows signals included in the information section 302 in time series tl, t2 . ．． ．． ．． ．． tn order, and the fifth
Figure B shows an example of the configuration of buckets (buckets) Pt edited separately for bands (1), (2), . . . (n).

The packet (bucket) Pt is transferred from the packet network 50 to the receiving terminal 52, and the packet assembling section 56 of the receiving terminal 52 performs the opposite process to that of the packet assembling section 54. More specifically, the reception bucket) Pt is decomposed into two fields, a header 300 and an information part 302, and the information part 302 is further decomposed into ADPCM codes Irl(k) to Irn(k) for each band. It is output to the ADPCM code interpolator 57.

FIG. 2 shows an embodiment of the ADPCM code interpolator 57 according to the subband encoding method. In this example, the code interpolator 57 consists of a low-band ADPCM code interpolator (1) of l units and a plurality of (1-1) units of high-band ADPCM code interpolators (2 to n). The high-frequency ADPCM code interpolators (2) to (n) may have the same configuration, and these
It is connected in multiple ways to the CM code interpolator (2).

The low-band ADPCM code interpolator (1) has an ADPCM code switching unit 11, which receives the low-band ADPCM code Irl(k) from the packet assembly unit 56 at the input end. The updated ADPCM code 1p(k) is also input to the code switching unit 11 from the Pauffer memory 12. The code switching unit 11 normally receives the low-band ADPCM code I rl (
k) from the output terminal 121 as the interpolated output code I rc
subpan):' output to the ADPCM decoder 58 as l(k). The code switching unit 11 has a signal terminal 110,
When the packet discard generation signal L(k) is input from now on, the ADPCM code generation signal Ip(k) from the buffer memory 12 is used instead of the code Irl(k) at the input terminal 111.
This is a selection circuit that outputs as an output code I rcl(k). As a result, interpolation of the packet discard section is performed.

The buffer memory 12 stores the ADP output from the switching unit 11.
The CM code Irl(k) is fed back, and the eightfold data I
The BUFFI is output to the pitch period estimator 13 in FIFO operation. The pitch period estimation unit 13 uses this signal IBUFF
When I is input, the pitch period is specified and a pitch period estimated value PIT indicating the value is outputted.

This pitch period estimated value PIT is input to the bar and family memory pointer designation section 14, and the same designation section 14 inputs the estimated value PIT.
(d) An address designating the corresponding storage location in the buffer memory 12 is output to the memory 12. The buffer memory 12 always outputs the interpolation data I pi (k) stored in the storage location specified by this address to the code switching unit 11 .

On the other hand, the high-band ADPCM code interpolators (2) to (n) are
The low-band ADPCM code interpolator (1) is connected as shown in the figure, and similarly, the ADPCM code switching unit 16, buffer memory 17, and buffer memory pointer designation unit 1 are connected to the low-band ADPCM code interpolator (1).
It has 8. High-band ADPCM code interpolator (2) to (n)
Then, the period estimation value PIT is sent from the pitch period estimating section 13 of the low-frequency interpolator (1) to the buffer memory pointer specifying section 18.
The buffer memory 17 always inputs the ADPCM code I p2(k
) ~I pn(k) is output to the switching unit 16.

Buffer memory 12 of low-band ADPCM code interpolator (1)
has a storage capacity M, and as described above, the ADPCM code Ircl(k) output from the ADPCM code switching unit 11 is input and its contents are updated sequentially. Equation (4) represents this state. The updated data I pi (k) is used for packet interpolation, and is subjected to decoding processing according to equation (5) in the subpan I'' ADPCM decoder 58.ADPC
The M code becomes a residual signal, and the voice band signal is restored from this residual signal.

IEUFFI(+e) = IBUFFI(m-1)
2≦m≦M (4) IBUFFI(+) = Q-”
(Ircl(k)) (5) FIG. 6A shows the residual signal and the packet discard section. The pitch period estimating section 13 constantly calculates the pitch period PIT by self-calculation processing and outputs it to the buffer memory 12, as shown in FIG. 6B. The buffer memory 12 always outputs the PIT data shown in FIG. 6C. The ADPCM code switching unit 11 responds to the signal L(k) and inputs I rl(k).
By selecting the same I pi (k) instead of , and using this data, the discarded section is compensated for.

The autocorrelation function R(t) of the pitch period designation unit 13 is (6)
Defined by Eq.

TWIN≦t≦TMAX where THIN and T WAX are the minimum and maximum number of samples, respectively, for which a pitch period is estimated to exist. For example, THIN=24, TMAX
= 128 ni, tri N = 12fz: Setting is complete. In this case, the size of buffer memory data IBUFFI is 2
It is 56. The value of t at which this autocorrelation function T(t) has a maximum value becomes the pitch period estimated value PIT.

The buffer memory pointer designation unit 14 uses this estimated value PT
Using T, the ADPCM code generation signal 1pl (k) is (
7) Calculate using the formula.

I pl(k)=Q(IBUFFI(PIT))
(7) Here, Q() indicates encoding processing.

The operation of the low-band ADPCM code interpolator (1) is as described above, and is basically the same as the operation of the full panto ADPCM code interpolator (8). The operation of the high-frequency ADPCM code interpolators (2) to (n) is similar to that of the low-frequency code interpolator (1).
), there is no equivalent to the pitch period estimating section 13 provided in
The value of M code interpolator (1) is used. Other operations are
The operation is substantially the same as that of the low-band ADPCM code interpolator (1). In this embodiment, the pitch period estimated value P
By using IT in both low and high bands, interpolation processing for high band ADPCM codes can be performed with only simple processing, thereby reducing the amount of processing.

(Effects of the Invention) According to the present invention, a packet transmission terminal generates one multiplexed packet whose information part is ADPCllf codes of a plurality of bands that occur within the same time period, and thereby transmits it to a packet network. There is one type of packet, and sub/
The hend ADPCM transmitting and receiving system is simplified. Furthermore, in the plural ADPCM code interpolators of the receiving terminal,
A pitch period estimator that outputs a signal pitch period to compensate for packet discards is provided only in the low-frequency interpolator, and this is shared by other high-frequency interpolators, resulting in a relatively complex arithmetic circuit. Only one pitch period estimator is required as in the full-pant ADPCM system. Therefore, high-precision subband ADPC can be performed without increasing the amount of processing.
Packet transmission using the M method is realized.

[Brief explanation of the drawing]

FIG. 1 is a relay system diagram showing an embodiment of a packet communication system to which the packet interpolation method according to the present invention is applied. FIG. 2 is a configuration of an ArJPCM code interpolator in a packet receiving terminal included in the embodiment shown in FIG. FIG. 3 is a functional block diagram showing an example of the configuration of the packet assembling unit in the packet transmitting terminal of the same embodiment. FIGS. 4A and 4B are the subband 'ADPCH encoder in the same embodiment. FIG. 5A and FIG. 5B are explanatory diagrams showing an example of the configuration of a multiplexed packet in the same embodiment. FIG. 6A is the relationship between the residual signal and the path-2 discard section. FIG. 6B is an explanatory diagram of pitch estimation using an autocorrelation function. FIG. 6C is an explanatory diagram of residual signal interpolation using estimated pitch. FIG. 7 is a packet communication system using a conventional full-pant ADPCM code. Figure 8A is a functional block diagram illustrating the configuration of a conventional full-pant ADPGN encoder, and Figure 8B is a configuration of a conventional full-pant ADPCM decoder. FIG. 9 is a functional block diagram similar to FIG. 8A showing an example, and FIG. 9 is a functional block diagram showing an example of the configuration of a conventional full-pan FADPCM code interpolator. 51. 52゜53゜54゜56゜57゜58゜Explanation of the code of the wife part Packet sending terminal Packet receiving terminal Subband ADPCII encoder Packet assembly section Packet assembly section ADPflll: M code interpolator Subband ADPCM decoder Patent applicant Oki Electric Industrial Co., Ltd. Su1 Basid ADPCM Bucket) 1τ Shizute A 1st
IBIJFF, Yu, ↑ IBIJFF,) *,! Total 1 pagno) 712-f ward YAn example'#/) A times self 1st eye 1i11Vl Nel: Yorobino + Chisashidairo BWJ no? 'T Nodo (Pt) / 'Structurer Diarrhea-j 5th A concave? Ri'ri (Pt)'s relative example to the 5th B episode) 1 riwa' open r Acri 1st) C episode 7's full bath > FADPCM bucket) Iyo of the recombination system 7th fluvanide ADPCM after issue summer r case 8th AWJ 8f 7) After band ADPCM + fog autopsy 8th BVJ

Claims (1)

[Claims] 1. In a packet interpolation method for interpolating packet discards in a packet communication system in which packets containing voice band signals encoded in ADPCM codes are transferred from a transmitting side to a receiving side via a packet network, the transmitting side comprises: divides the audio band signal into a plurality of frequency bands within the same period and encodes them into corresponding ADPCM codes of the plurality of bands, stores the ADPCM codes of the plurality of bands within the same period in an information section, and stores the ADPCM codes of the plurality of bands within the same period; a header is added to the multiplexed packet to assemble it into a multiplexed packet, and the multiplexed packet is transmitted to the receiving side through the packet network, and when the receiving side receives the multiplexed packet, the multiplexed packet is sent to the ADPCM of the plurality of bands. decompose into codes, temporarily store the decomposed ADPCM codes, and if there is no packet discard, select the decomposed ADPCM codes of multiple bands;
When a packet is discarded, the accumulated ADPCM
A packet interpolation method characterized in that a code is selected for a predetermined period, the selected ADPCM codes of a plurality of bands are decoded and synthesized, and the audio band signal is output. 2. In the method according to claim 1, the information part of the multiplexed packet includes ADPC information within the same period of the plurality of bands.
A packet interpolation method characterized in that M codes are arranged by band and by time series. 3. Encode the audio band signal into an ADPCM code, and
In a packet transmitting device that assembles PCM codes into packets and transmits them to a receiving side through a packet network, the device divides the voice band signal into multiple frequency bands within the same period and encodes them into ADPCM codes of multiple bands. subband ADPCM encoding means; and packet assembling means for storing ADPCM codes of a plurality of bands within the same period in an information section, adding a header to the information section, assembling a multiplexed packet, and transmitting the multiplexed packet to the packet network. and on the receiving side, upon receiving the multiplexed packet, the multiplexed packet is decomposed into the ADPCM codes of the plurality of bands, the decomposed ADPCM codes are temporarily stored, and if no packet is discarded, the Decomposed multi-band ADPCM
When a code is selected and a packet is discarded, the stored ADPCM code is selected for a predetermined period, the selected ADPCM codes of multiple bands are decoded and combined, and the voice band signal is output. A packet transmitting device characterized by: 4. On the transmitting side, the packet containing the audio band signal encoded in the ADPCM code is received through the packet network, and
In a packet receiving device that decodes a DPCM code and outputs the voice band signal, on the transmitting side, the voice band signal is divided into a plurality of frequency bands within the same period and encoded into ADPCM codes of corresponding multiple bands. ADPCM codes of multiple bands within the same period are stored in an information section, a header is added to the information section and assembled into a multiplexed packet, and the multiplexed packet is transmitted to the device through the packet network, The apparatus includes: upon receiving the multiplexed packet, a packet decomposition unit that decomposes the multiplexed packet into ADPCM codes of the plurality of bands; temporarily storing the decomposed ADPCM codes;
ADPCM code interpolation means that selects ADPCM codes of multiple bands output from the decomposition means if there is no packet discard, and selects the stored ADPCM codes for a predetermined period when there is packet discard; A subband ADPCM that decodes and synthesizes the ADPCM codes of multiple bands and outputs the audio band signal.
A packet receiving device comprising: decoding means. 5. The apparatus according to claim 4, wherein the ADPCM code interpolation means
It includes a plurality of ADPCM code interpolators corresponding to CM codes, and a relatively low-frequency interpolator among the plurality of ADPCM code interpolators is configured to perform a relatively low-frequency interpolator among the plurality of ADPCM codes inputted from the decomposition means. a first ADPCM code switching means that always outputs a low-frequency code and outputs a first interpolation ADPCM code when a packet discard generation signal is input; and an ADPCM output from the first ADPCM code switching means. In addition to accumulating the codes for a predetermined period and outputting them in the order of input,
a first buffer means for outputting the ADPCM code stored at a designated address as a first interpolation ADPCM code; and a pitch cycle of the ADPCM codes output by the first buffer means in the order of input is estimated, and pitch period estimating means for outputting an estimated value of the period; and first addressing means for specifying an address corresponding to the estimated value to the first buffer means; Generally, the high-frequency interpolator always outputs a relatively high-frequency code among the ADPCM codes of multiple bands inputted from the decomposition means, and when a packet discard occurrence signal is inputted, it performs the second interpolation. ADPCM code output from the second ADPCM code switching means is stored for a predetermined period, and the ADPCM code stored at the specified address is transferred to the second ADPCM code. A second buffer means for outputting as an interpolation ADPCM code, and a second addressing means connected to the pitch period estimating means and designating an address corresponding to the estimated value to the second buffer means. Characteristic packet receiving device.