基于Sigmoid学习率BP神经网络的非特定人语音识别算法研究  

Research on Non-specific Speech Recognition Algorithm Based on Sigmoid Learning Rate BP Neural Network

作　　者：戚龙[1] 

机构地区：[1]吉林化工学院信息与控制工程学院,吉林132022

出　　处：《价值工程》2017年第33期200-202,共3页Value Engineering

摘　　要：本课题基于非特定人语音识别的原理和过程,结合BP神经网络的建模理论及特点,主要研究了BP神经网络模型在语音模式识别中的应用问题。同时针对标准BP神经网络训练收敛速度慢及容易陷入局部最小的缺点,提出了Sigmoid学习率BP神经网络训练算法,并通过仿真计算,得出在非特定人语音识别应用方面Sigmoid学习率BP神经网络算法的收敛速度要优于标准BP神经网络算法且识别率更高。Based on the principle and process of non-specific speech recognition, this paper focuses on the application of BP neural network model in speech pattern recognition, combined with the modeling theory and characteristics of BP neural network. At the same time,the Sigmoid learning rate BP neural network training algorithm is proposed for the training of BP neural network, and the Sigmoid learning rate BP neural network is obtained by simulating the non-specific speech recognition application. The convergence rate of the algorithm is better than that of the standard BP neural network algorithm and the recognition rate is higher.

关 键 词：BP神经网络 学习率 语音识别 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]