基于SOC FPGA的车载语音识别系统设计

Design of SOC FPGA based on board speech recognition system

针对车载语音识别系统对实时性要求高、小型化及低功耗等要求,提出了一种基于SOC FPGA的语音识别系统的设计方案。该系统采用软硬件结合的设计思路,语音信号采集及预处理部分由FPGA的Verilog DHL实现;特征参数提取及模式匹配算法由ARM核处理器在内核Linux系统上用C语言实现。硬件板级测试表明,该系统识别速度分别为DSP系统的4.59倍及嵌入式ARM系统的1.96倍。

Aiming at the high demand of real-time performance and miniaturization and low power consumption of speech recognition system in vehicle environment,an on-board voice recognition system is designed based on SOC FPGA.The system is designed by combination of software and hardware.Speech signal acquisition and preprocessing part realized by VDHL of FPGA.Characteristic parameters extraction and pattern matching algorithm are implemented by C language in the Linux kernel of ARM.Board-level hardware tests show that the system＇s recognition speed is 4.59 times of DSP system and 1.96 times of the embedded ARM system separately.