JPS6476100A - Voice compressor - Google Patents

Abstract

PURPOSE: To double the performance of compression without changing a conventional encoding part by compressing a useless part (the repeated part of approximately the same waveform in silence or a voiced sound) in a voice. CONSTITUTION: When the contents of a recording buffer memory 2 are data indicating silence, code data are not written in a memory 7 and the data length of silence is recorded in a synthesis control memory 4. At the time of inputting a voiced sound, the similarity of stored data between buffers is judged so that data for one pitch are stored in one buffer, only representative data are stored in the memory 7, the other buffer data are not stored and only the number of buffers having similar data is stored. At the time of reproducing data, code data for specified data length are read out from the memory 7 in accordance with information read out from the memory 4 and sent to a decoder 9 repeatedly by the specified number of times. Consequently, an inputted voice is recorded by a compression rate extremely larger than the compression rate of a conventional encoder.