Arrhythmic segmentator of voiced sound in speech signals

摘要

The segmentator marks points in time signals, for each word. For example, the start and end points of high energy voiced sounds are set as synchronisation time points. These are read out when training the system to learn words and are used to determine the segment boundaries. The segment lengths correspond to the sound lengths. The lengths of voiced segments are preferably corrected to a whole number multiple of the longest period length which can be recognised in the signal.