written 6.7 years ago by | • modified 6.5 years ago |
Subject: Speech Processing
Topic: LPC and Parametric Speech Coding
Difficulty: Low
written 6.7 years ago by | • modified 6.5 years ago |
Subject: Speech Processing
Topic: LPC and Parametric Speech Coding
Difficulty: Low
written 6.5 years ago by |
(i) Channel vocoders: Channel vocoder models the vocal tract as bank of band-pass filters. The analysis filter at the transmitter side finds the average square value of energy for input speech 50 times per second.
(ii) Basically, channel vocoder try to match the frequency profile of the input speech signal.
(iii) They do not reproduce the speech sample.
(iv) It can be seen that not all frequency components speech signal are important. The vocal tract resonant at formants frequency. Formant vocoders are developed to use the frequency.
(v) The range of formant frequency known so the formants are estimated along with bandwidth from a speech segment using a bank of filters designed to match the range of formants.
(vi) At the receiver end, the excitation signal-either a periodic pulse train is used to excite the tunable filter that are tuned to $ R_x $.