基于自适应的小波阈值和GTEO进行说话人识别

小波压缩是基于小波系数的阈值的一个简单去噪方法.它对于所有的含噪语音,都使用一致的阈值,不仅压缩了噪声,也压缩了部分语音成分,因此滤掉的语音感知质量会受到极大的影响.在小波去噪过程中采用了自适应阈值小波包方法.同时把小波去噪和推广的TEO结合起来,去提高系数的鲁棒性.为了进一步提高识别率,在识别阶段,采用改进的MCE算法.实验结果显示,提出的方法取得了较好的效果.

Speaker Recognition Based on Adapted Wavelet Threshold and GTEO

Wavelet shrinkage is a simple de-noising technique based on the threshold of the wavelet coefficients.The estimated threshold is supposed to define the limit between the wavelet coefficients of the noise and those of the target signal.Applying threshold uniformly to all wavelet coefficients not only suppresses additional noise but also some speech components.consequently,the perceptive quality of the filtered speech will be greatly affected.Adapted wavelet packet threshold during wavelet de-noising process is appliled.At the same time,wavelet de-noising is combined with GTEO(Generalized Teager Energy Arithmetic)in order to enhance the system robustness.In order to further improve recognition rate,at the recognition stage,modified MCE arithmetic is applied.Simulation results showed that the proposed system leads to better performance.