嵌入式DSP声纹识别系统的设计与实现

The Design of Speech Recognition System based on Embedded DSP

本系统采用美国模拟器件公司的高性能Blackfin嵌入式数字信号处理器,完成对语音信号实时采集,语音采集完成后,通过对其整体样本进行快速傅立叶变换得到语音样本的频谱,然后再通过多级带通数字滤波器,提取不同频段下的频率幅值,作为声纹识别的特征值,完成对语音的辨别。实验表明,该设计对特定人语音的识别率达到了80%,有一定实用价值。

Design of the United States through the company offers high-performance Blackfin embedded digital signal processor support completed the acquisition of real-time voice. Upon completion of the acquisition, the overall sample was fast Fourier transform voice samples of the spectrum, then through the multi-stage band pass filter, band different from the amplitude frequency, so as to identify the eigenvalue completion of the voice identification. Experiments show that the design of specific voice has recognition rate of 80%, the results have been satisfactory and it has been confirmed some practical value.