论文部分内容阅读
在过去30年中,针对许多基于事件的语音基音检测算法进行了研究。小波变换表示法为分析信号的不连续性提供了很好的框架。但是,对于离散情况,因低语音信号分辨率而出现了时间分辨率问题。本文提出了基于二进小波变换的语音基音算法。该算法充分利用了二进小波变换的平移不变形和各尺度上小波系数间的相关性,有效改善了去噪过程中噪声放大问题。提供的实例表明,该算法的去噪效果较佳。
In the past 30 years, many event-based speech pitch detection algorithms have been studied. Wavelet transform representation provides a good framework for the analysis of signal discontinuities. However, for discrete cases, temporal resolution issues arise due to the low speech signal resolution. This paper proposes a speech pitch algorithm based on binary wavelet transform. The algorithm takes full advantage of the translational invariance of the binary wavelet transform and the correlation between the wavelet coefficients at each scale, and effectively improves the noise amplification problem in the denoising process. The provided examples show that the algorithm has better denoising effect.