作者: Michele Cavazza , Alberto Ciaramella
DOI: 10.1121/1.399368
关键词:
摘要: The device obtains several characteristic parameters from a standard sentence said by speaker and compares them with average of the same stored in an internal memory previously calculated. According to comparison, it probability value that spoken belongs threshold normalized parameter variance calculation circuit. If is overcome, considers verified. A circuit determines real instants beginning end using noise-adaptive order limit between these two time interval over which are be aligns as just calculated reference sentence, obtaining lengths sounds composing spoken. variable controlled deviations histogram vectors.