Speech Plot Display for the Deaf-mute based on Combined Characters Encoding of Speech Signal

摘要

This paper describes a novel speech plot display approach that extracts combined time-frequency-energy feature as a new feature vector to present speech signal. The combined feature is displayed on CRT by plot patterns and the deaf-mute can make use of their own brain to identify speech for training their oral ability effectively. Firstly, speech signal undergoes a series of preprocessing course. Secondly, we make use of a filter group to process spectrum analysis for speech signal. The output of each band-pass filter generates a time-frequency-energy value. Thirdly, we calculate that the feature value through band-pass filter lies in which place in full array and obtain the combined feature value. Finally, we generate speech plot through coordinates partition, area computation and coordinates calculation of combined feature value.