Research And Implementation On The Methods Of Extracting Glottal Wave Based On Inverse Filtering

Voice plays an important role in our daily life. It is not only the most effective communicating tool, but also has the function of emotional expression and artistic creation. Voice disorder not only affects people’s normal communication, but also influences the physical and mental health. The voice source detection is the core content of voice rehabilitation, so the research on voice source detection has very important theoretical significance and clinical value. This article mainly focuses on inverse filtering to extract glottal wave from voice.This article adopts liner prediction(LP) inverse filtering(IF) to extract glottal wave which is applied to voice detection and then optimize the algorithm by experiments.The quality of the glottal wave will be the best under the conditions of using 1200 points of the vowel /a/ as analysis voice, choosing the LP coefficients of the closed glottal phase as the inverse filtering transfer function, setting the predictor order 16. On the basis of the LP inverse filtering, this paper proposed a method based on weighted liner prediction(WLP) inverse filtering to extract glottal wave. The quality of the glottal wave based on WLP is better than LP when the length of the weighted function is 24. By analyzing the inverse filtering glottal wave with electroglottography, this paper draws a conclusion that the IF glottal wave can reflect vocal fold movement to a certain extent. Therefore, inverse filtering glottal wave has a certain theoretical significance and clinical value.The article will be to replace the electroglottography, and then provide theoretical support for the development of the Chinese language function testing and rehabilitation system based on digital tract simulation.