Programs for Digital Speech Processing
PICOLA: Improved time scale modification for
Audio signals.
PICOLA is detailed here.
PICOLA is a modified version of Time Domain Harmonic Scaling (TDHS).
The name "PICOLA" is abbreviation of "Pointer Interval Controlled
OverLap and ADD". In precise meaning, PICOLA is NOT Harmonic Scaling
BUT time scale modification for audio signals with periodicity. The
original algorithm was developed by MORITA Naotaka in 1986.
- Source Code(picola.c : 7 KBytes)
- Feathers
- Selectable Auto-correlation or AMDF method for detecting
fundamental frequency or pitch period
(some modifications should be required)
- Including compansion rate error feed back control
- Supporting signed 16 bit, 8000 kHz sampling, and single channel
raw audio format only.
VSELP : Speech Codec for Japanese Digital
Cellular Phone
- Archive(vselp.tar.gz : 42KB)
- Document (in Japanese)
- Feathers
- Satisfying RCR STD-27B
- Including error control encoding/decoding (Hard Viterbi)
- Supporting signed 16 bit, 8000 kHz sampling, and single channel
raw audio format only.
SS: Speech Enhancement by Spectral Subtraction
- Archive
- Feathers
- learning noise spectra without pause detection
- Note: Supporting signed 16 bit and single channel
raw audio format only.
FFT: Fast Fourier Transform