抄録
We describe the development of a new kind of speech analysis and representation tool,"Speech-ART".The tool presents various parameter-trajectories and several selected frame-parameters together as graphs on fivewindows on a display screen.One of its main features is that users can intuitively select an appropriate analysisframe based on the clear color images(previously developed by us)of phoneme sequences in speech.Moreover,this tool is also suitable for segmenting phoneme sequences from connected speech signals,which is an importantstage in training models used in speech recognition systems,neural networks,and so on.
