SPEECH SECTION DETECTION DEVICE

• Main Authors: SHITAYA MITSUO, KANEKO KAZUMA, MORIHIRO YOSHIHARU
• Format: Patent
• Language: English
• Published: 30.04.1993
• Subjects: ACOUSTICS; MUSICAL INSTRUMENTS; PHYSICS; SPEECH ANALYSIS OR SYNTHESIS; SPEECH OR AUDIO CODING OR DECODING; SPEECH OR VOICE PROCESSING; SPEECH RECOGNITION

PURPOSE:To detect a correct speech section by updating a noise standard pattern only when a voiceless state is decided by voiced/voiceless decision making and the distance between patterns is smaller than a threshold value for noise standard pattern update. CONSTITUTION:A noise standard pattern update decision part 12 is equipped with a noise update distance threshold value memory 13 stored with the inter- pattern distance threshold value for noise update to be used and then the noise standard pattern update decision part 12 compares the inter-pattern distance outputted by an inter-pattern distance calculation part 8 with this noise update inter-pattern distance threshold value to decide the update. Namely, when the voiced/voiceless decision part 9 decides that the decision result of a speech section decision part 11 indicates a point before the start of a voiced section and the noise standard pattern update decision part 12 decides that the distance between the feature of an input signal and the standard pattern is less than the noise update inter-pattern distance threshold value set smaller than the voiced/voiceless decision inter-pattern threshold value, a noise standard pattern calculation part 6 updates the noise standard pattern.