摘要
研究BP神经网络技术在数字语音识别中的应用,以基于语音信号产生的数字模型作为突破口,对所采集到的语音信号进行预处理,提取Mel频率倒谱系数,并将特征参数序列进行非线性时间规整为固定的帧数以便于BP神经网络的训练和识别。由MATLAB的实验数据分析可得,基于BP神经网络的数字语音识别技术具有很高的实用价值、数字语音识别率高。
The BP neural network technology in the application of digital speech recognition,based on the figures of speech signalmodel as a Breakthrough, Collected for the speech signal preprocessing, The extraction of Mel frequency cepstrum coefficient,andwill feature parameters for nonlinear time sequence neat for the fixed frame is advantageous for the BP neural network of trainingand recognition.By the MATLAB analysis of experimental data available,digital speech recognition based on BP neural networkhas a high practical value,digital speech recognition rate is high.
出处
《电脑知识与技术（过刊）》
2015年第7X期141-142,共2页
Computer Knowledge and Technology
基金
长江大学工程技术学院科学研究发展基金资助(15j0401)
