对语音信号进行预处理,包括预滤波、采样、量化、分帧、加窗以及端点检测等内容:采用能够反映人对语音信号的感知特性的Mel频率倒谱系数作为语音信号的特征参数:应用目前在语音识别中广泛应用的技术——动态时间规整技术作为识别算法。测试结果表明,利用该技术进行语音识别,准确率可达到95%以上,识别效果良好。 Preprocess the speech signal, including prefiltering, sampling, quantization, framing, windowing and endpoint detection. Using Mel frequency cepstrum coefficients that reflect the perceived characteristics of speech signals as the characteristic parameters of speech signal: Currently widely used in speech recognition technology - dynamic time warping technology as a recognition algorithm. The test results show that using this technique for speech recognition, the accuracy rate can reach more than 95%, and the recognition effect is good.