Design of real-time voice speed change system based on TMS320C5409

LPC algorithm

LPC coding regards the speech signal s(n) as the output signal of a time-varying vocal tract system under the excitation of the glottis. For voiced sounds, the glottis excitation is a periodic pulse train; for unvoiced sounds, the glottis excitation is a random noise sequence. The simplified block diagram of the speech signal is shown in Figure 1.

In LPC coding, a frame of speech signal can be represented by simplified model parameters, such as voiced and unvoiced discrimination, pitch period, gain G, and digital filter coefficient {a1}, so that 3kb/s coding can be obtained. During decoding, in LPC coding, voiced sound can be regarded as the excitation of a pulse train with a period, where the pulse period is the pitch period. Therefore, in order to change the speech rate, we lengthen the frame length of the speech, and then add several pulse train excitations to obtain a variable-speed speech signal.

Hardware system design

This system is a complete application system, which mainly completes the real-time LPC encoding and decoding of voice signals and voice speed change. In addition, it also has keyboard control function and voice recording and playback function. Therefore, this system not only needs external SRAM to store voice data, but also needs a keyboard control device connected to DSP. This system is mainly composed of TMS320C5409, TLC320AD50 , SST39VF400A and CY7C133, and the system block diagram is shown in Figure 2. The 32K word on-chip RAM provided by TMS320C5409 is used to store real-time running programs and data, and the off-chip SRAM (CY7C133) is used to store encoded voice data. The working process of the system is as follows: After the system is powered on, the TMS320C5409 is reset through the reset function. After TMS320C5409 is reset, the internally fixed self-boot program moves the program and initialization data on the Flash (SST39VF400A) to the on-chip RAM. Then TMS320C5409 executes operations according to the control of the keyboard, runs LPC encoding operations (recording), LPC decoding and voice speed change (playback), and adjusts the voice speed change rate.

TMS320C5409 is used to implement the voice encoding and decoding algorithm, and TLC320AD50 is used to complete the D/A and A/D conversion functions of the voice signal. TMS320C5409 and TLC320AD50 communicate through the McBSP serial port. In order to reduce the burden on the CPU, this system is designed to communicate with the CPU by interrupting the CPU after receiving or sending a group of units through DMA. The interface between TMS320C5409 and TLC320AD50 is shown in Figure 3.

Adopt DMA mode, that is, every time the serial port receives or sends a group of units, it will automatically trigger DMA to move them to an internal buffer, and when the buffer is full, it will tell the CPU to process it through interruption. At this time, it is best to use ABU (Auto Buffering) mode for DMA, which can effectively prevent the data in the buffer from being washed away by new data when the serial port rate is high. When changing the voice speed, you only need to change the frame length of the 13MA channel accordingly.