摘要
针对说话人语音特征空间边界存在模糊性的特点,构建了一种量子神经网络识别分类器,用于说话人识别,以改善存在交叉数据的语音特征参数的分类效果。提出了一种基于人工免疫算法的量子间隔训练方法,以改善传统量子神经网络训练算法的不足。以TIMIT语音库为测试语音,与传统BP网络和基于常规梯度下降量子间隔训练算法的量子神经网络做对比实验。实验证明,算法能有效提高说话人识别系统的识别率,同时与高斯混合模型相比,具有更好的抗噪声性能。
To tackle the problem that fuzziness exists in the speech feature space boundary, a speaker rec- ognition model based on quantum neural network (QIqN) was constructed to improve the classification per- formance of speech feature parameters. One method based on the artificial immune algorithm was proposed to the training quantum neural network. The experimental results show that the performance of the pro- posed method is better than that of the traditional BP neural network and gradient descent training QNN, and also more robust to noise than GMM.
出处
《解放军理工大学学报(自然科学版)》
EI
北大核心
2012年第3期242-246,共5页
Journal of PLA University of Science and Technology(Natural Science Edition)
基金
江苏省自然科学基金资助项目(BK2009059)
关键词
量子神经网络
说话人识别
人工免疫算法
多层传递函数
高斯混合模型
QNN (quantum neural network) ~ speaker recognition~ artificial immune algorithm
multileveltransfer functiong GMM(Gaussian mixture models)
作者简介
王金明(1972-),男,博士,副教授,研究方向:语音信号处理、说话人识别,E-mail:wjm_ice@163.com