SPEECH ANALYZING AND SYNTHESIZING METHOD

摘要

PURPOSE:To improve the quality of a restored speech by increasing quantization efficiency. CONSTITUTION:A speech analysis part extracts LSP(line spectrum couple) parameters and sound source parameters from an input speech message by extraction parts 3 and 4. When the parameter extraction from the whole message ends, a table generation part 22 generates an encoding table and a decoding table according to the distribution characteristics of the LSP parameters. An encoding part 5 encodes the extracted LSP parameters and sound source parameters by using the encoding table. The encoded message and decoding table are stored in a storage part 23. A speech synthesis part reads the decoding table corresponding to a speech message to be outputted out of the storage part 23 and writes it in a storage part 10, and then reads a encoded parameter series out in order next, and a decoding part 9 decodes it by referring to the decoding table in the storage part 10 and sends the decoded parameters to a sound source generation part 11 and a vocal tract filter part 12, thereby synthesizing the speech.