hi all, sorry to disturb you again with some silly questions, but i really hit a big rock here . 1) the vocal tract of speech production can be modeled by LPC synthesis filter, right ? it is shown by the spectral magnitude form of LPC coefficient filter that looks like a spectral envelope of the speech signal. am i right ? now, the problem is how about the phase response ? i have tried thousand times that although the LPC coeff really models the spectral envelope of speech signal, but the phase is totally different, so the waveform of reconstructed signal will be totally different from the original. Do you guys think it is okey ? i mean i read in Deller discrete time processing of speech signal that the human ears can not sense the phase different between two speech signal (phase deaf). But how come the sound of the reconstructed speech is very buzzy and low intelegibility ? 2) the glottal filter, is it really necessary to use ? how necessary ? and what could possibly be wrong if i'm not use it rather i use just a simple periodic pulse ? thanks in advance Eris Ristemena ============ DSP Laboratory Telkom School of Technology Bandung, Indonesia 62-22-7564108 ext. 2325 http://students.stttelkom.ac.id/~eris_ristemena/ ============ |
|
need help about LPC 10
Started by ●February 28, 2000
Reply by ●February 28, 20012001-02-28
I am looking for binary data (words or phrases), decoded in lpc-10 and usable on TI's obsolete TMS5220 chips. do you know any source or do you have some predecoded data? I have a few including a disk from TI with the dumps of 8 preprogrammed VM61.... Voice-memories. BTW: are there any tools on the market to record voice and save in lpc-10? I am a hardware-developing engineer and work on TI's DSP cards. If anyone has an application for such a voice-recoder using a TMS320C... please let me know. I remember a demoboard for TMS32010 from TI including a codec (in the late 80's...), but cannot find this item anymore at our business... Thank you in advance, Michael Becker Mannheim Germany |