基于MFCC系数的语音交互系统设计

Design of speech interaction system based on MFCC coefficient

现有的语音交互系统在提取特征参数时难以保证滤波器的通道数量,导致音频识别准确率较差,为提高音频的识别能力,基于MFCC系数设计语音交互系统。在硬件设计中,设计通讯寄存器芯片架构,设计音频录入并行电路,保证大量音频数据在短时间内录入的稳定性。在软件设计中,将音频信号预加重,量化信号输入信噪比,基于MFCC系数提取音频数据特征参数,保证滤波信号通道数量,设计音频交互算法,计算训练最大似然度。在实验中对比四种语音交互系统的音频识别能力,实验数据显示,该系统音频数据识别准确率在五类音频信号中约为67%-97%,其他三种系统的识别准确率分别为62%-92%、66%-89%、67%-91%,由此可见该系统识别准确率高于这三种系统,基于MFCC系数的语音交互系统可以更好地处理与转换音频数据。

The existing voice interaction system is difficult to ensure the number of channels of the filter when extracting feature parameters,resulting in poor audio recognition accuracy.In order to improve the audio recognition ability,the voice interaction system is designed based on MFCC coefficient.In the hardware design,the communication register chip architecture and audio input parallel circuit are designed to ensure the stability of a large number of audio data input in a short time.In the software design,the audio signal is pre weighted,the signal input signal-to-noise ratio is quantified,the characteristic parameters of audio data are extracted based on MFCC coefficients to ensure the number of filtered signal channels,the audio interaction algorithm is designed,and the maximum likelihood of training is calculated.In the experiment,the audio recognition capabilities of the four voice interaction systems are compared.The experimental data show that the audio data recognition accuracy of the system is about 67%-97%in the five types of audio signals,and the recognition accuracy of the other three systems are 62%-92%,66%-89%and 67%-91%respectively.It can be seen that the recognition accuracy of the system is higher than that of the three systems,The voice interaction system based on MFCC coefficient can better process and convert audio data.