Abstract： A spcetrogram is a grey scale image,which represents the energy changes of a speech signal. Auto-matic segmentation is an initial phase in the acoustic一phonetic analysis of automatic speech recognitionbased on spectrograms. Speech segmentation can be defined as the process of dividing the spectrograminto a sequence of segments，each segment indicating phonemic characteristics. This paper presents amethod of automatic segmentation with image processing techniques. We describe two special functionswhich indicate the intensity changes of the spectrograms called. Together with these two functions,weused adaptive threshold techniques to detect the location of the edges for each segment. The thresholdwas calculated based on an optimum relation equation which was defined using interpolating linear nulti-ple regression. After the preliminary segmentation,a segmentation check procedure was taken to checkthe segmentation results. The algorithm was evaluated by comparing the automatic segmentation resultwith another segmentation result carried out by a phonetic expert. This automatic segmentation facilityis a part of an automatic feature extraction program appiled in a speech analysis system.
潘凌 云 孙达传 吴美朝. 语音识别中基于语谱图 的语音音素分割方法[J]. 浙江大学学报（理学版）, 1997, 24(1): 43-46.
Pan L YSun D Cwu M c. A Method of Automatic Segmentation for Speech Recognition Based on Spectrograms. Journal of ZheJIang University(Science Edition), 1997, 24(1): 43-46.