基于EMD法的语音信号特征提取

Feature Extraction of Speech Signal Based on Empirical Mode Decomposition

特征提取是目标识别的关键,如何从有限的测量数据中获取有效、可靠的特征参数,是特征提取中重点考虑的问题。本文采用EMD方法对语音信号进行频率特征提取,可以较好地降低语音信号的冗余度,实验结果表明:EMD方法是处理非平稳信号的有效方法,它运用于语音信号特征提取是可行的。

Feature extraction is the key to object recognition. How to obtain effective, reliable characteristic parameter from the limited measured data is a question of great importance in feature extraction. In this paper, EMD is used to extract frequency feature of a speech signal, which can better reduce its redundancy. EMD is an effective method for processing non-stationary signal. It is feasible to extract frequency feature of speech signals with EMD.