JP3362829B2 - Cell conversion method for multiple data - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to data generated in bursts such as voice data compressed at a low bit rate and having a plurality of different burst periods and burst lengths, such as IP packets and ATM cells. The present invention relates to a method of converting a cell made up of a set of data.

[0002]

2. Description of the Related Art ATM (Asynchronous Transfer Mode) positioned as a basic network technology for multimedia communication
Irrespective of the nature of the transfer data such as the type of transfer speed,
One of its features is that it can process all information such as images, sentences, and voices on the same physical line. Therefore, it is considered that the ATM network, which is currently used mainly for data communication, is transformed into a data + voice network which also has a function of cost reduction of ATM-related equipment and inevitably also processing of real-time voice information. Against this background, there is an urgent need to develop voice communication technology on ATM.

Particularly in a private line for a specific purpose such as between private networks or between a base station and a switch in mobile communication, the limited network capacity is effectively utilized,
In order to reduce the cost per line, low bit rate compression of voice information and silence compression that does not transmit silence are essential. The bit rate according to the compression method widely used at present is in the range of 32 to 4 kb / s.
When converting voice information having a bit rate lower than kb / s into cells, a delay for accumulating one cell of voice information of one channel (celling delay) generally increases with a decrease in bit rate, so that one cell It is necessary to consider a method to reduce cellization delay without reducing cell transfer efficiency by storing voice information packets of multiple channels with short user information length.

In the recent voice compression algorithm of 8 kb / s or less, since the compression process is performed in frame units of a plurality of voice samples, the compressed voice to be output is obtained in bursts in tens to hundreds of bits unit for each frame. To be Therefore, by packing the compressed voice for each frame as a voice information packet in a standard cell, unnecessary cell formation delay can be suppressed and highly efficient voice information can be transmitted.
However, in order to maintain high data transfer efficiency, a certain number of line multiplexing numbers are required, so it is necessary to realize low bit rate voice cell conversion and multiplexing processing in one set.

That is, in the low bit rate voice encoding system of 8 kb / s or less, as shown in FIG. 6, data is output in bursts in units of tens to hundreds of bits for each frame rate. For example, as shown in FIG. 7, channels ch-1 to c
Each of the low bit rate audio data of h-3 is packetized, for example, with a header h added for each frame, and the audio packets are AT-ordered in the order of occurrence in these three channels.
The payload of the M standard cell is sequentially packed to form a cell with multiplexed voice information. In the case of silence compression, voice packets are not output in silence frames. In this way, cellization delay can be suppressed and highly efficient voice information can be obtained.

[0006] Such low bit rate audio information is multiplexed.
The configuration shown in Fig. 8 is considered as the structure for liquefying and filling the cells.
To be C that supports various encoding methods for the audio compression unit 11
ODEC (codec) 12a ₁~ 12a_m, 12b₁
~ 12b_n, 12c₁~ 12c_pIs provided, CODE
C12a₁~ 12a_m, 12b₁~ 12b_n, 12c₁
~ 12c_pEach has a frame cycle of Ta, Tb, Tc
And the number of bits in one frame is also the pulse width in the figure.
Are different from each other.

CODECs 12a _{1 to} 12a _m , 1 of the same kind
2b ₁ ~12b _n, 12c ₁ per ~12c _p, voice input and output portion 14a of the multiplexing process CLAD unit 13, 14b, 1
4c and the voice input / output unit multiplexes bursts (compressed voice information) of a plurality of channels (lines) input from a plurality of CODECs, that is, associates each multiplexing time slot with a channel. And multiplex.

The voice input / output units 14a, 14b, 14c are connected to the cell decellizing units 15a, 15b, 15c, respectively, and the voice information multiplexed by the voice input / output unit 14a is voice packetized for each channel. The voice packets are packed in the cells in the order in which they are generated, and ATM cells are created. Similarly, the multiplexed voice information of the voice input / output units 14b and 14c is similarly converted into the cell decellization units 15b and 15c, respectively.
To make an ATM cell. These cell deceller 15
The ATM cells generated in a, 15b and 15c are sent to the ATM network 16.

A input from the ATM network 16
The TM cell is supplied to the corresponding one of the cell decellizing units 15a, 15b, 15c by the coding method of the voice information in the cell and is decellized, that is, separated into compressed voice for each frame of each channel. It is supposed to be multiplexed voice,
It is supplied to the corresponding one of the audio input / output units 14a, 14b, 14c, separated for each channel by the audio input / output unit, and supplied to the corresponding CODEC of the audio compression unit 11. Each CODEC of the audio compression unit 11 is connected to any of the 64 kb / s audio lines 18 via the selective connection unit 17. In other words, the CODEC of the same type as the CODEC of the other party
Is selectively connected by the selective connection unit 17.

[0010]

Since the frame period and the number of bits of compressed information (burst) (burst length) differ depending on the encoding system, the channel voices of the same encoding system are multiplexed and processed. . In other words, in the conventional multiplexing technology, the time lengths of the time slots corresponding to the channels are the same, and the time slots and the burst length of the compressed voice assigned to this time slot are made to match, and channels with different burst lengths must be multiplexed. I can't. Further, even when the frame period is different, the correspondence between each channel and the time slot number is not constant, and multiplexing cannot be performed. That is, conventionally, it was necessary to provide a voice input / output unit for each encoding method.

Also, in the cell assembling unit, since the cells are packed without gaps, if the voice packets have different lengths, the packing process becomes complicated. Since decelling is performed, it is necessary to provide a cell decelling unit for each coding method.
As described above, since the voice input / output unit and the cell decelling unit are provided for each encoding system, the device scale becomes large and the cost becomes high. When introducing a new CODEC or changing the CODEC to be used. However, it is necessary to newly introduce a voice input / output unit and a cell decelling unit to support the CODEC, and the voice multi-processing CLAD unit 13 itself is changed, which causes a problem of high cost. there were.

These problems occur not only in the case of converting compressed voice into ATM cells but also in the case of converting into IP packets (packets in a packet transmission network). Further, not only when the compressed voice is multiplexed into cells (ATM cells, IP packets, etc.), but also when a plurality of burst-generated data are generally cellized, the burst cycle and burst length are different. Has similar problems.

[0013]

According to the present invention, the greatest common divisor of burst periods of a plurality of input data is selected as a basic period, and the basic period is divided into a plurality of time slots.
For each input data, time slots are sequentially assigned to fill the burst length, and if there is transmission data in the assigned input data in the order of time slots, at least 1.
Fill cells in bursts.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment in which the present invention is applied to a case where a plurality of compressed voices are multiplexed and converted into a cell is shown in FIG. 1, and portions corresponding to those in FIG.
In this embodiment, all the CODECs 12a ₁ , 12a ₂ , 12b ₁ , 12 are irrespective of the encoding format in the audio compression unit 11.
c _1, ... is connected to the multiplexing demultiplexing unit 21 is connected to one voice input and output portion 14 of the multiplexing demultiplexing unit 21 and the multiplexing process CLAD unit 13. The voice input / output unit 14 is connected to one cell decelling unit 15.

In the present invention, all input data (channels), that is, all CODECs 12a ₁ , 12a ₂ , 12b.
Compressed audio output data Da ₁₁ , D for ₁ , 12c ₁ , ...
_{a 12, ..., Da 21,} Da 22, ..., Db 11, Db 12, ...,
The burst periods Ta, Tb, Tc, of Dc ₁₁ , Dc ₁₂ , ...
_Let the greatest common divisor of ... Be the basic period T _R. As shown in FIG. 2A, in this embodiment, the basic period T _R is equally divided by a value which is larger than and close to the maximum burst length (number of bits) Ba, Bb, Bc, ... Of all input data (channels). To form time slots T _S1 , T _S2 , T _S3 , ..., T _Sn . Figure 6
In the example shown in, the frame period is 10ms, 20ms, 3
Since it is 0 ms and 40 ms, the basic period T _R may be set to 10 ms, and the time slot length may be set to 189 bits or more capable of accommodating the maximum burst length.

Bursts of input data (channels) are sequentially assigned to the time slots T _{S1 to} T _Sn . The input data (channels) are the first channels Da ₁₁ , Da ₁₂ , ..., The second channels Db ₁₁ , Db ₁₂ , ..., The third channels Dc ₁₁ ,.
Dc ₁₂ , ..., And the burst periods Ta, Tb, Tc of these three input data are Ta = T _R , Tb = 3 T.
FIG. 2B shows the state of allocation to each time slot when _R , Tc = 4 T _R , and the burst length is Ba, Bb, and Bc bits (all of Ba, Bb, and Bc are smaller than the bit length Bt of the time slot). Show. That is, the first channel is used as the first time slot T _S1 and the second channel is used as the second time slot T _S2.
, The third channel is assigned to the third time slot T _S3 , respectively. If there is transmission data in the assigned input data (channel), the burst is sequentially inserted into each time slot T _S1 , T _S2 , .... First basic period T
The first channel burst Da ₁₁ in the first slot T _S1 of _R1 .
, The second channel burst Db ₁₁ is inserted into the second slot T _S2 , and the third channel burst Dc ₁₁ is inserted into the third slot T _S3 . In the second basic cycle T _R2 , the burst Da ₁₂ of the first channel is inserted in the first slot T _S1 , but it is not inserted in the second and third channels because there is no transmission burst. Since the burst period Ta of the first channel is equal to the fundamental period T _R, the fundamental period T _R1, T
Burst D of the first channel in the first slot T _S1 of _R2 , ...
_{a 11, Da 12, Da 13} , ... but are respectively inserted, since the burst period Tb of the second channel is a 3T _R, since the transmission data is not generated only every 3T _R, 4 th fundamental period T _R4 The burst Db ₁₂ is inserted in the second slot T _S2 of the second channel and the burst Dc ₁₂ of the third channel is inserted in the third slot T _S3 of the fifth basic cycle T _R5 . Similarly, the bursts of each channel are inserted in the assigned time slots. As shown in FIG. 2B, the insertion of each burst is performed so that the start of the time slot and the start of the burst coincide with each other.

In this way, the multiplexing / demultiplexing unit 21 time-division-multiplexes the input data by inserting a transmission burst into a predetermined time slot having a basic cycle as a frame. This multiplexed compressed voice information is sent to the voice input / output unit 14 for the correspondence between the basic period and the time slot position,
The relationship between each time slot and the number of bits of the compressed voice information (burst) inserted therein is recognized, and each burst and its time slot identifier ID are output to the cell assembling unit 15 as a set.

FIG. 3 shows a specific functional configuration example of the voice input / output unit 14. The multiplexed compressed audio information is sequentially input to the input buffer 23, this transfer clock is counted by the transfer bit number counter 24, and this counter 24 is reset every basic cycle T _R (basic frame synchronization signal), whereby the basic frame Of the data transfer clock of each channel is counted by the counter 24, and the count value is compared with the number of time slot bits B _S stored in advance in the definition information storage unit 25 to determine the basic cycle T. The number of time slots between _R is identified. In addition, the frame synchronization signal of each CODEC is also input,
For example, FIG. 2C shows the frame synchronization signal for high level during the burst length in each slot and low level during the burst length in the first basic period T _R1 of FIG. 2B. . The number of output bits per frame of CODECa is 1 for each of Ba, CODECb, and c.
If the output bit numbers per frame are Bb and Bc, respectively, the burst length is equal to these bit numbers.
The level is low when there is no CODEC output (silence). Since the frame synchronization signal of each CODEC is input as described above, the difference in the count value of the transfer bit number counter 24 at the time of the rise and the fall of the frame synchronization signal of this CODEC is detected to detect each time slot. The number of effective bits is detected (recognized). Therefore, it is possible to take out only the number of effective bits for each time slot from the input buffer 23, and output this voice data (burst) and the number ID of the time slot as a set to the cell decellizing unit 15. The recognition is performed by the recognition unit 26. Further, it is possible to confirm whether or not voice data of the corresponding channel is present by the frame synchronization signal of each CODEC, and perform voice / silent discrimination in the case of silence control.

By such processing, it is possible to provide an interface for transmitting / receiving a set of voice data in bit unit and time slot number ID to / from the cell decellizing unit 15 by asynchronous transfer. Also, the cell decellization unit 1
5, it is possible to limit the channel difference to only the packet length (burst length), it is possible to operate at a high speed with a unique clock speed, and it is possible to absorb the processing complexity associated with multiplex processing.

Next, a concrete example of the functional configuration of the cell assembling unit 15 is shown in FIG. 4 and its operation will be described. Since the CODEC used for each voice signal is selected for each call processing, the correspondence information between each time slot number and the type of CODEC used, that is, the burst length is stored in the storage unit 31, and the voice input / output unit 14 The voice data is written in the cell assembly memory 32 under the control of the packet writing controller 33. Time slot number ID corresponding to the voice data
Is input from the voice input / output unit 14, and the burst length selection unit 3
A corresponding burst length in the storage unit 31 is selected by 4, and this is given to the packet writing control unit 33, and the voice data is written in the cell assembly memory 32 by the burst length. In this way, the voice data of each time slot is written in the cell assembly memory 32.

When the cells are filled in the order in which they are written in the memory 32, that is, in each burst (frame) of each channel, the standard cell read control unit 35 writes them in the order in which they are written in the memory 32. Each burst is read out, and a header h such as the channel ID and the burst length generated by the packet header generation unit 36 is added to the burst header, and the packet is packed into the cell as in the case shown in FIG. 7 as a voice packet. When the writing is completed, the standard cell header H generated by the standard cell header generation unit 37 is added to this to make an ATM cell, which is written in the ATM output FIFO 38. Every time the writing of the audio data corresponding to each time slot into the memory 32 is completed, the data transfer control unit 39 sends a transfer control signal to the audio input / output unit 14 to request the transfer of the next audio data.

If the frame period of voice data (channel) is short, a plurality of continuous bursts may be assembled into one voice packet within the range where voice delay does not matter. In the cell disassembly in the cell disassembly unit 15, the standard cell header analysis unit 42 analyzes the header H of the ATM cell input to the ATM input FIFO 41, and further the packet header analysis unit 43 analyzes the header h of each voice packet. , Each time slot is identified, and written into the cell disassembly memory 44 for each burst under the control of the write control unit 45, and from the cell disassembly memory 44 to the read control unit 46, the multiplexed compressed audio is controlled. The data is read, assembled, and transferred to the voice input / output unit 14. The voice input / output unit 14 temporarily stores the voice data from the cell decellization unit 15 in the output buffer 47 (FIG. 3) and then supplies it to the multiplexing / demultiplexing unit 21.

It is conceivable that a FAX signal and other types of information on the voice line are converted into cells by the same multiprocessing CLAD unit. In this case, for example, as shown in FIG. 5, as the voice input / output unit 14, compressed voice input / output ports 51, 6 are provided.
A 4 kb / s audio input / output port 52 and an input / output port 53 for other purposes are provided, and the same compressed audio input / output port 51 as the audio input / output unit shown in FIG. 3 is used.
32 kb / s for G3 and 6 for G4 for AX signals
Since a band of 4 kb / s is required, after detecting the presence / absence of a FAX signal, if it is FAX, a 64 kb / s voice having an interface with the cell decelling unit 15 at a frame cycle of 125 μs of a 64 kb / s base. I / O port 52 for data, and if it is data, the I / O port 53 for data has an interface with the cell deceller 15 for other purposes.
To use. These three input / output ports 51, 52, 53
Is selected by the priority control and data transfer control unit 54. In this way, different types of information can be handled without changing the interface between the voice input / output unit 14 and the cell decelling unit 15.

[0024] In the example shown in FIG. 2, T _R = 10 ms, when the data rate _{t r = 2 (Mb / s} ), the multiplexing number (number of time slots) in n is 32 to 64, bits of the time slot the number B _S is secured 640-320 bits, a frame period is compression rate of a multiple of 10 ms 8 kb / s to
It is applicable to various CODECs of about 4 kb / s. In FIG. 2B, the basic period T _R is equal to or larger than the maximum burst length and is equally divided into values close to the maximum burst length to form time slots. However, this is not always necessary, and for burst lengths longer than the time slot length B _S. Alternatively, a plurality of consecutive time slots may be assigned.

In the above, the present invention is converted into the ATM cell by compressing and multiplexing the voice, but the present invention is applicable not only to the ATM cell but also to the IP packet. Further, the present invention can be applied to the case where a plurality of data having different burst periods and different burst lengths are multiplexed and converted into cells.

[0026]

As described above, according to the present invention, by selecting the basic period, dividing it into time slots, and allocating these to each channel (input data), the burst period and burst length of the input data can be obtained. Is changed,
Even if a new one is added, it is possible to perform multiple cell conversion by the common cell decellization unit without providing a cell decellization unit for each burst cycle, which simplifies the overall hardware configuration. It can be made at low cost.

[Brief description of drawings]

FIG. 1 is a block diagram showing an example of a cell conversion device to which an embodiment of the present invention is applied.

FIG. 2A is a diagram showing a relationship between a basic cycle and a time slot, B is a basic cycle and an example of allocation of each data of the time slot, and C is a diagram showing an example of a CODEC frame synchronization signal.

3 is a block diagram showing a specific functional configuration example of a voice input / output unit 14 in FIG.

FIG. 4 is a block diagram showing a specific functional configuration example of a cell decellization unit 15 in FIG. 1.

FIG. 5: Cellizing section decelling section 1 for multiple types of information
5 is a block diagram showing an example of a functional configuration of a voice input / output unit 14 for activating No.

FIG. 6 is a diagram showing a bit rate, a frame rate, and the number of bits of one frame in various encoding systems.

FIG. 7 is a diagram for explaining a proposed method of assembling a multiplexed cell of compressed voice of a plurality of channels.

FIG. 8 is a block diagram showing a configuration for implementing a proposed method for assembling and disassembling a compressed voice for multiplexing cells.

─────────────────────────────────────────────────── ─── Continued Front Page (58) Fields surveyed (Int.Cl. ⁷ , DB name) H04L 12/56 300

Claims (5)

(57) [Claims] 1. A plurality of burst data are input,
These data have at least one different burst cycle and / or burst length, and in the cell conversion method for converting these data into cells each consisting of a predetermined block of data, the greatest common divisor of the burst cycle of each input data is Is selected as the basic cycle, this basic cycle is divided into multiple time slots, and for each input data, time slots are sequentially allocated as long as the burst length is satisfied. A cell conversion method for a plurality of data, characterized in that if there is transmitted data, the cells are packed in at least one burst at a time. 2. The method for converting cells of a plurality of data according to claim 1, wherein the length of the time slot is not less than the maximum burst length in the input data. 3. The plurality of cells according to claim 1 or 2, wherein the filling of the input data into the cell is performed without a gap by adding an identifier of the data and a filling length for each filling. Data cell conversion method. 4. The allocation is performed for each burst of the input data to multiplex a plurality of input data, and each time slot is identified for each basic period in synchronization with the multiplexed burst train. 4. The cell conversion method for plural data according to claim 3, wherein the identifier and the burst are used as a set in the cell packing process. 5. A filling length according to the identifier is detected,
5. The cell conversion method for a plurality of data cells according to claim 4, wherein cell packing is performed every time the added value of the burst of the corresponding time slot in the multiplexed burst sequence becomes this packing length.