SPEECH RATE CONVERSION WITHOUT EXTENSION OF INPUT DATA DURATION, USING SPEECH INTERVAL DETECTION

摘要

When a delivered speed of a listening speech(speech speedy is slowed down, a connection ordergenerator (8) always monitors a data length of inputspeech, an output data length calculated previously bya conversion function concerning a preset scalingfactor, and a data length of actual output speech inpredetermined processing unit, then decides connectionorder so as not to cause inconsistency among them. Thespeech data and the connection data are connectedwithout omission of speech information by controllinga speech data connector (9). When power of an inputsignal data is calculated to discriminate a speechinterval and a non-speech interval, a threshold valuefor power is decided according to a maximum value ofthe power and the difference between the maximum valueand a minimum value.