论文部分内容阅读
采用二阶差分耳蜗模型对语音信号进行特征参数提取 ,获得了基于听觉谱的语音识别前端特征参数 ,同时根据听觉谱特征提出了一种“幅和频差积”距离测度 ,识别算法采用端点放松两帧 ,路径斜率限制在 1/ 2到 2之间的改进型 DTW算法 .在小词汇量非特定人 (SI)的识别环境下 ,计算机模拟结果表明此法在对 0～ 9十个数字以及小词汇量的 SI识别时 ,其正识率可达 98%以上 ,且具有较好的鲁棒性
Second-order differential cochlear model was used to extract the feature parameters of speech signals, and the front-end speech recognition parameters based on auditory spectrum were obtained. At the same time, a distance measure of amplitude and frequency difference product was proposed according to auditory spectrum features. The recognition algorithm adopted endpoint relaxation Two-frame DTW algorithm with the path slope limited from 1/2 to 2. Under the environment of small vocabulary unspecified (SI) identification, the computer simulation results show that this method has better performance in 0-9 digits and Small vocabulary SI recognition, the positive rate of more than 98%, and has good robustness