论文部分内容阅读
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。
Wavelet compression is a simple denoising method based on the threshold of wavelet coefficients. It uses a consistent threshold for all noisy speech, not only compressing the noise but also compressing a portion of the speech component so that the quality of the filtered speech perception can be greatly affected. In wavelet denoising process, an adaptive threshold wavelet packet method is adopted. At the same time, wavelet denoising and popularizing TEO are combined to improve the robustness of the coefficients. In order to further improve the recognition rate, an improved MCE algorithm is adopted in the recognition stage. The experimental results show that the proposed method has achieved good results.