基于快速神经网络算法的非特定人语音识别

提出一种用于语音识别的改进的快速神经网络算法 ,即动态不等步长的误差分段学习算法。将步长看作误差和网络节点输出的函数 ,对各权值按不同步长进行动态调整 ,并将其应用于一个基于前馈神经网络模型的非特定人语音识别系统。实验表明 ,该算法比传统 BP算法在训练速度上可提高十几倍 ,训练出的语音识别网络系统具有较高的识别率

Speaker-independent speech recognition based on a fast NN algorithm

An improved learning algorithm--dynamic different step error segmenting algorithm is presented, in which the step is regarded as the function of the error and the output function of network node, and weight is regulated dynamically by different step. By adopting the fast NN algorithm, a speaker-independent speech recognition system based on a BP NN is set up. The experiment shows that the new algorithm is over 10 times faster than the traditional BP algorithm and the resulting neural network has better performance and spreading ability.