CN1236631C - Vocoder unit for mobile communication system and its phonetic frame displayching method - Google Patents

Abstract

The present invention relates to a vocoder device in a mobile communication system and a phonetic frame dispatching method of the vocoder device. The device comprises a main control CPU, n digital signal processors DSP, a time division multiplexing TDM interface and an interruption generating logical unit, wherein the main control CPU is used for receiving encoded phonetic frames, writing the received phonetic frames into vocoder units of corresponding digital signal processors DSP for decoding and simultaneously reading the encoded phonetic frames of the vocoder units of the corresponding digital signal processors DSP; each of the n digital signal processors DSP, can realize m vocoder units and simultaneously complete the function of coding and decoding m paths of phonemes; the time division multiplexing TDM interface is arranged in DSP for receiving and transmitting phonetic PCM code streams; the interruption generating logical unit is used for generating timer interruption signals which are transmitted to the main control CPU and the digital signal processors DSP. The method mainly processes the phonetic frames of all vocoder units in one interruption to make the main control CPU not frequently respond to interrupt processing any longer, and the performance of the main control CPU is increased.

Description

Vocoder device and voice frame scheduling method in mobile communication system

technical field

The invention relates to a vocoder device, in particular to a vocoder device in the field of mobile communication and a voice frame scheduling method thereof.

Background technique

In the mobile communication system, in order to effectively utilize the spectrum resources of the air interface, the speech transmitted between the mobile station MS (Mobile Station) and the base station is a speech frame compressed by speech coding. The vocoder is a device that realizes conversion between speech frames compressed by speech coding and standard PCM codes. General speech coding algorithms have the characteristics of large amount of calculation and strong real-time performance. Digital Signal Processor (DSP, Digital Signal Processor) has a very fast computing speed and flexible implementation, which is very suitable for application in the vocoder system. In the vocoder system, a main control CPU (Central Processing Unit) is required to manage the DSP at the same time, and the main control CPU needs to take out the encoded speech frames from the DSP or store the speech frames that need to be decoded into the DSP. In the Chinese patent document with publication number CN1409575A (publication date: April 9, 2003), a vocoder system using multi-port routing communication and a service frame exchange method thereof are provided. Its system composition includes: a main control CPU, used to receive the business frame sent by the base station, write into the multi-port routing communication unit, and send the business frame of the multi-port routing communication unit to the base station; the multi-port routing communication unit is used to store The business frame is divided into n main control CPU sending/DSP receiving queues and n main control CPU receiving/DSP sending queues; n digital signal processors receive and send services according to the interrupt signal of the multi-port routing communication unit frame; an E1 interface connected to the mobile switching center; a clock receiving unit 1; and a clock receiving unit 2 for extracting the clock of the E1 interface. The business frame exchange method of the system is as follows: the main control CPU receives the business frame through the interface with the base station, writes the business frame into the multi-port routing communication unit, accesses the interrupt generation unit, interrupts the corresponding digital signal processor, and the digital signal processor reads Fetch the business frame and give it to the digital signal processor to process the corresponding one of the vocoder units realized by time division multiplexing. The business frame of the vocoder unit is written into the multi-port routing communication unit by the digital signal processor, and access The interrupt generation unit interrupts the main control CPU, and the main control CPU receives the service frame and sends it to the base station.

In this existing patent document, the system uses a multi-port routing communication unit as a buffer unit for voice frames between the main control CPU and the digital signal processor, resulting in a complex system structure, and the voice frames need to be stored and forwarded through the multi-port routing communication unit. The time delay of the voice frame is increased; at the same time, the system uses the interrupt method to access the voice frame, and each voice frame will trigger an interrupt. When there are many vocoder units in the system, the main control CPU will be frequently interrupted, resulting in The performance of the master CPU is degraded. When there are many vocoder units controlled by the main control CPU, if the main control CPU adopts the method of processing the voice frame of one vocoder unit each time interrupted, the main control CPU will waste a considerable amount of time in saving the environment for interrupt processing , reducing the efficiency of the master CPU. For example, in a code division multiple access system, the duration of a speech frame is 20 milliseconds. If a main control CPU controls 480 vocoder units, an interruption will be generated every 42 microseconds. Frequent interruptions will greatly reduce the performance of the main control unit. The operating efficiency of the CPU.

Contents of the invention

The technical problem to be solved by the present invention is: in order to overcome the disadvantages of complex hardware structure in the prior art and difficulty in adding vocoder units in the system, the purpose of the present invention is to provide a vocoder device in a mobile communication system, The hardware structure is simple, and the system vocoder unit is highly integrated.

Another object of the present invention is to provide a kind of voice frame scheduling method based on the buffer zone that DSP internal RAM forms, in an interrupt, the voice frames of all vocoder units are processed, so that the main control CPU no longer frequently responds Interrupt handling improves the performance of the master CPU.

The vocoder device of the present invention comprises the following components:

Main control CPU is used to receive the coded speech frame, and writes the received speech frame into the vocoder unit in the corresponding DSP to decode, and reads the coded speech frame of the vocoder unit in the corresponding DSP at the same time; n There are m vocoder units in each DSP, which can simultaneously complete the encoding and decoding functions of m channels of voice frames; a time division multiplexing (TDM, Time Division Multiplexing) interface is also included in the DSP for receiving and sending Voice PCM code stream; interrupt generating logic unit, used to generate timing interrupt signal and send it to main control CPU and DSP. The shared RAM composed of DSP internal RAM, that is, this part of RAM can be accessed not only by DSP, but also by the main control CPU through the interface with DSP. The shared RAM is further divided into sending buffers for storing voice frames to be decoded; receiving buffers for storing voice frames encoded by DSP; pointer buffers for storing operation pointers of sending buffers and receiving buffers.

The voice frame scheduling method of the vocoder device of the present invention:

The interrupt generation logic unit regularly generates interrupts and sends them to the main control CPU and DSP, and the interval of regular interrupts is equal to the duration of a voice frame.

After the main control CPU responds to the interrupt, it executes the following processing flow for all vocoder units:

201. Read the read operation pointer and the write operation pointer of the sending buffer from the pointer buffer, thereby calculating the free queue in the sending buffer.

202. Write the speech frame to be decoded into the corresponding queue of the sending buffer.

203. After resetting the write operation pointer of the send buffer, write back to the pointer buffer.

204. Read the read operation pointer and write operation pointer of the receiving buffer from the pointer buffer, thereby calculating the position of the voice frame in the receiving buffer.

205. Read the encoded speech frame from the corresponding queue in the receive buffer.

206. After resetting the read operation pointer of the receiving buffer, write back to the pointer buffer.

After the DSP responds to the interruption, perform the following processing flow for all vocoder units:

207. Read the read operation pointer and write operation pointer of the sending buffer from the pointer buffer, thereby calculating the position of the voice frame in the sending buffer.

208. Read out the speech frame to be decoded from the corresponding queue in the sending buffer, and send it to the corresponding vocoder unit for decoding and output.

209. After resetting the read operation pointer of the send buffer, write back to the pointer buffer.

210. Read the read operation pointer and the write operation pointer of the receiving buffer from the pointer buffer, thereby calculating the free queue in the receiving buffer.

211. Write the speech frame encoded by the vocoder unit into the idle queue of the receiving buffer.

212. After resetting the write operation pointer of the receiving buffer, write back to the pointer buffer.

Using the method and device of the present invention, compared with the prior art, the multi-port routing communication unit for buffering voice frames is removed, and the hardware structure is simple. In terms of interruption generation, the interval of timing interruption is the duration of a voice frame. The speech frames of all vocoder units are processed in the interrupt, instead of the speech frames of each vocoder unit being interrupted once, so that the main control CPU no longer frequently responds to interrupt processing, and the performance of the main control CPU is improved.

Description of drawings

Fig. 1 is a kind of vocoder device schematic diagram that the present invention proposes;

Fig. 2 is a schematic diagram of a voice frame scheduling method proposed by the present invention;

Fig. 3 is the sending buffer structural diagram of the first kind of specific implementation of the shared RAM in Fig. 1;

Fig. 4 is the receive buffer structural diagram of the first kind of specific implementation of the shared RAM in Fig. 1;

Fig. 5 is the pointer buffer structural diagram of the first kind of specific implementation of the shared RAM in Fig. 1;

Fig. 6 is the sending buffer structural diagram of the second kind of concrete implementation of the shared RAM in Fig. 1;

Fig. 7 is the receiving buffer structural diagram of the second kind of specific implementation of the shared RAM in Fig. 1;

FIG. 8 is a pointer buffer structure diagram of the second implementation of the shared RAM in FIG. 1 .

Detailed ways

Below in conjunction with accompanying drawing, the implementation of technical scheme is described in further detail:

As shown in FIG. 1 , the hardware part of the present invention is composed of a main control CPU 101 , a DSP 102 and an interrupt generation logic unit 103 . DSP 102 includes a TDM interface and internal RAM accessible by master CPU 101 .

The shared RAM inside the DSP102 is divided into three parts, which are the sending buffer queue 301 , the receiving buffer queue 302 and the pointer buffer queue 303 . Fig. 3 is the sending buffer structural diagram of the shared RAM in Fig. 1; Fig. 4 is the receiving buffer structural diagram of the shared RAM in Fig. 1; Fig. 5 is the pointer buffer structural diagram of the shared RAM in Fig. 1; What the district queue 301 deposits is that main control CPU101 writes in DSP102, waits for the voice frame that the vocoder unit decodes; What the receiving buffer queue 302 stores is that the vocoder unit of DSP completes encoding, waits for the voice frame that main control CPU101 takes out; Pointer buffer queue 303 deposits the pointer that master control CPU and DSP102 operate sending buffer queue 301 and receiving buffer queue 302, including: master control CPU101 to the write operation pointer (TxWrite Ptr) of sending buffer queue 301, master control CPU101 to the read operation pointer (Rx Read Ptr) of receiving buffer queue 302, DSP102 to the read operation pointer (Tx Read Ptr) of sending buffer queue 301, DSP102 to the write operation pointer (Rx Write Ptr) of receiving buffer queue 302 . In FIGS. 3, 4, and 5, CHANNEL refers to a queue of buffered speech frames corresponding to a vocoder unit, and each CHANNEL corresponds to a vocoder unit. Each DSP102 can implement M vocoder units to perform speech codec simultaneously. Each CHANNEL can buffer K speech frames for the vocoder unit.

After main control CPU101 receives voice frame, store in the buffer zone of main control CPU101 temporarily, wait for interrupt logic generation unit 103 to produce timing interrupt, the interval of timing interruption is equal to the time length of a voice frame, after interrupt generation, trigger main control CPU101 will temporarily The speech frame that saves is written in the sending buffer queue 301 corresponding to the corresponding vocoder unit of DSP102, contains control information in the speech frame, indicates the serial number of the destination DSP102 of this speech frame and the unit number of the vocoder unit to which it belongs. Control CPU101 and voice frame is written into the process of sending buffer formation 301 as follows: main control CPU101 visits pointer buffer formation 303 at first, reads the write operation pointer (Tx Write Ptr) of main control CPU101 of corresponding CHANNEL and the read operation pointer of DSP102 (Tx ReadPtr), the main control CPU 101 compares these two pointers, checks whether the corresponding CHANNEL of the sending buffer queue 301 is free, if not idle, then reports to the police and exits, if the corresponding CHANNEL of the sending buffer queue is idle, then writes Voice frame, then main control CPU101 is written back pointer buffer 303 again to the write operation pointer (Tx Write Ptr) of sending buffer queue 301 and adds 1. Master control CPU101 reads the process of voice frame from receiving buffer queue 302 as follows: master control CPU101 visits pointer buffer queue 303 at first, reads the write operation pointer (Rx Read Ptr) of master control CPU101 of corresponding CHANNEL and DSP102 Operation pointer (Rx Write Ptr), main control CPU101 compares these two pointers, check whether the corresponding CHANNEL of receiving buffer formation 302 has voice frame, if there is voice frame, then read from the corresponding CHANNEL of receiving buffer formation 302 Get the voice frame, then add 1 to the read operation pointer (RxRead Ptr) of the main control CPU101 and write back to the pointer buffer 303.

After the interrupt generation logic unit 103 triggers the DSP102 to generate an interrupt, the DSP102 first polls the sending buffer queue 301, and sends the speech frame to be decoded to the vocoder unit for decoding, and the decoded PCM code stream is output from the TDM interface, and simultaneously the output from the TDM The PCM code stream input by the interface is encoded into a voice frame and written into the receiving buffer queue 302 . DSP102 reads the process of voice frame from sending buffer queue 301 as follows: DSP102 visits pointer buffer queue 303 at first, reads the write operation pointer (TxWrite Ptr) of the main control CPU101 of corresponding CHANNEL and the read operation pointer (Tx Read Ptr) of DSP102 Ptr), DSP102 compares these two pointers, check whether the corresponding CHANNEL of sending buffer queue 301 has a voice frame, if there is a voice frame, then read the voice frame from the corresponding CHANNEL of sending buffer queue 301, then DSP102 The read operation pointer (Tx Read Ptr) plus 1 is written back to the pointer buffer 303. DSP102 writes the process of speech frame in receiving buffer queue 302 as follows: DSP102 visits pointer buffer queue 303 at first, reads the read operation pointer (Rx Read Ptr) of the main control CPU101 of corresponding CHANNEL and the write operation pointer (RxWrite) of DSP102 Ptr), DSP102 compares these two pointers, checks whether the corresponding CHANNEL of receiving buffer queue 302 is idle, if idle, then voice frame is written in the corresponding CHANNEL of receiving buffer queue 302, writes DSP102 then Operation pointer (Rx Write Ptr) adds 1 and writes back pointer buffer 303.

The main difference between Fig. 6, 7, 8 and Fig. 3, 4, 5 is that all vocoder units in DSP102 share a sending buffer formation 301 ' and a receiving buffer formation 302 ', and in Fig. 3, 4, 5 Each vocoder unit has a separate queue in transmit buffer queue 301 and receive buffer queue 302 . Generally, the structures shown in Figures 3, 4, and 5 require more RAM space than the structures shown in Figures 6, 7, and 8. Using the structures shown in Figures 3, 4, and 5, each CHANNEL corresponds to a vocoder unit, and CHANNEL is the private queue of the vocoder unit, and it is impossible for other vocoder units to access the voice of the vocoder unit Frame data, the main control CPU101 and DSP102 can determine that this speech frame belongs to a certain vocoder unit when reading a speech frame. And adopt the structure shown in Fig. 6, 7, 8, all vocoder units share the same queue, the speech frame stored in the queue must include control information, indicate the vocoder unit that this speech frame belongs to, the main control When CPU101 and DSP102 read a voice frame, they cannot determine the ownership of the voice frame. Only after analyzing the control information contained in the voice frame can they determine the vocoder unit to which the voice frame belongs.

In Fig. 6, 7, 8, the process that main control CPU101 writes voice frame into sending buffer formation 301 ' is as follows: main control CPU101 visits pointer buffer formation 303 ' at first, reads the writing operation pointer (Tx of main control CPU101 Write Ptr) and the read operation pointer (Tx Read Ptr) of DSP102, the main control CPU101 compares these two pointers, checks whether the sending buffer queue 301' is free, if not idle, then gives an alarm and exits, if the sending buffer queue has Idle, then write voice frame, then main control CPU101 adds the number of writing voice frame to the write operation pointer (Tx Write Ptr) of sending buffer queue 301 and writes back pointer buffer 303 ' again. Master control CPU101 reads the process of voice frame from receiving buffer queue 302 ' as follows: master control CPU101 visits pointer buffer queue 303 ' at first, reads the write operation of the read operation pointer (Rx Read Ptr) of master control CPU101 and DSP102 Pointer (Rx Write Ptr), main control CPU101 compares these two pointers, checks whether receiving buffer formation 302 ' has voice frame, if there is voice frame, then reads voice frame from receiving buffer formation 302 ', then The reading operation pointer (Rx Read Ptr) of main control CPU101 adds the number of read voice frame and writes back pointer buffer 303 '.

DSP102 reads the process of voice frame from sending buffer queue 301 ' as follows: DSP102 visits pointer buffer queue 303 ' at first, reads the write operation pointer (Tx Write Ptr) of main control CPU101 and the read operation pointer (Tx Read Ptr) of DSP102 Ptr), DSP102 compares these two pointers, checks whether there is a voice frame in the sending buffer queue 301 ', if there is a voice frame, then reads the voice frame from the sending buffer queue 301 ', and then sets the read operation pointer of DSP102 (Tx Read Ptr) adds the number of voice frames read and writes back the pointer buffer 303'. DSP102 writes voice frame into the process of receiving buffer queue 302 ' as follows: DSP102 at first visits pointer buffer queue 303 ', reads the read operation pointer (Rx Read Ptr) of main control CPU101 and the write operation pointer (Rx Write Ptr) of DSP102 Ptr), DSP102 compares these two pointers, checks whether the receive buffer queue 302' is free, if idle, then writes the voice frame into the receive buffer queue 302', and then writes the write operation pointer (RxWri te Ptr) plus the number of voice frames written into the pointer buffer 303 '.

Claims (8)

1. A vocoder device in a mobile communication system, comprising the following components: The main control CPU is used to receive the encoded speech frame, and write the received speech frame into the vocoder unit in the corresponding digital signal processor DSP for decoding, and at the same time read the vocoder unit in the corresponding digital signal processor DSP Speech frame after unit encoding; There are n digital signal processors DSP, and there are m vocoder units in each digital signal processor DSP, which can simultaneously complete the encoding and decoding functions of m channels of speech frames; a time-division multiplexing TDM interface is also included in the DSP for Receive and send voice PCM code stream; The interrupt generation logic unit is used to generate a timing interrupt signal and send it to the main control CPU and the digital signal processor DSP. 2. The vocoder device in the mobile communication system according to claim 1, wherein the RAM inside the digital signal processor (DSP) constitutes a shared RAM. 3. the vocoder device in the mobile communication system as claimed in claim 2, is characterized in that, described shared RAM is divided into send buffer, receive buffer, pointer buffer; Wherein send buffer, be used for depositing The decoded speech frame; the receiving buffer, used to store the speech frame encoded by the digital signal processor DSP; the pointer buffer, used to store the operation pointers of the sending buffer and the receiving buffer. 4. the vocoder device in the mobile communication system as claimed in claim 3 is characterized in that, described operation pointer comprises main control CPU to the write operation pointer of sending buffer, and main control CPU is to the read operation of receiving buffer Pointer, DSP read operation pointer to send buffer, DSP write operation pointer to receive buffer. 5. A voice frame scheduling method of the vocoder device as claimed in claim 1, characterized in that the method comprises the following steps: (1) After the main control CPU receives the voice frame, it waits for an interrupt to generate a logic unit to generate a timing interrupt; (2) The interrupt generation logic unit generates an interrupt and sends it to the main control CPU and the digital signal processor DSP; (3) After the main control CPU responds to the interrupt, the voice frame is written into the corresponding sending buffer of the corresponding vocoder unit of the digital signal processor DSP, and the main control CPU reads the voice frame from the receiving buffer; (4) After the digital signal processor DSP responds to the interruption, it reads the speech frame from the sending buffer, sends the speech frame to be decoded to the vocoder unit for decoding, and outputs the decoded PCM code stream from the TDM interface, and at the same time sends the speech frame from the TDM The PCM code stream input by the interface is encoded into voice frames and written into the receiving buffer. 6. the speech frame scheduling method of vocoder device as claimed in claim 5, it is characterized in that, step (3) also comprises: read the read operation pointer and the write operation pointer of sending buffer from pointer buffer, to sending The read operation pointer of the buffer zone is compared with the write operation pointer to determine whether the corresponding CHANNEL is free to write voice frames, thus calculating the idle queue in the send buffer, and adding 1 to the write pointer of the send buffer, Write back to the pointer buffer; read the read operation pointer and write operation pointer of the receiving buffer from the pointer buffer, and compare the read operation pointer and write operation pointer of the receiving buffer to determine whether there is a need to read in the corresponding CHANNEL voice frame, thus calculate the position of the voice frame in the receiving buffer, add 1 to the read operation pointer of the receiving buffer, and write it back to the pointer buffer. 7. the speech frame scheduling method of vocoder device as claimed in claim 5, it is characterized in that, step (4) also comprises: read the read operation pointer and the write operation pointer of sending buffer from pointer buffer, to sending The read operation pointer of the buffer zone is compared with the write operation pointer to determine whether the corresponding CHANNEL has a voice frame, thereby calculating the position of the voice frame in the sending buffer, adding 1 to the read operation pointer of the sending buffer, and then writing Back to the pointer buffer; read the read operation pointer and write operation pointer of the receiving buffer from the pointer buffer, and compare the read operation pointer and write operation pointer of the receiving buffer to determine whether there is any need to read in the corresponding CHANNEL Voice frame, thus calculate the idle queue in the receiving buffer, add 1 to the write operation pointer of the receiving buffer, and write back to the pointer buffer. 8. the speech frame scheduling method of vocoder device as claimed in claim 5 is characterized in that, the interval of described timing interruption is the duration of a speech frame, during an interruption, the speech of all vocoder units frame is processed.

Images