论文部分内容阅读
为模拟基底膜对语音信号的分析，本文提出了一种类似小波变换的时频分析方法。该方法克服了ＳＦＴ（短时Ｆｏｕｒｉｅｒ变换）分析对高、低频段具有相同的时间分辨率和频率分辨率的缺点，弥补了小波变换只能粗糙地模拟基底膜带通滤波器特性的不足。在识别阶段，本文采用了多模糊状态综合处理的识别模式，该模式有利于对各频段统计特征参数的提取和加强。实验表明，该模型具有良好的鲁棒性，较好地模拟了人的听觉系统对语音的识别过程。
To simulate the voice of the basement membrane analysis, this paper presents a time-frequency analysis similar to wavelet transform. This method overcomes the shortcomings of the SFT (short-time Fourier transform) analysis on the high-frequency and low-frequency bands with the same time resolution and frequency resolution, and makes up for the shortcoming that the wavelet transform can only coarsely simulate the characteristics of the basement membrane band-pass filter. In the recognition stage, this paper adopts the identification mode of multi-fuzzy state comprehensive processing, which is beneficial to the extraction and strengthening of the statistical characteristic parameters in each frequency band. Experiments show that this model has good robustness and well simulates the human speech recognition process.