celp_codec

This is a simple CELP voice codec implementation. http://en.wikipedia.org/wiki/Code-excited_linear_prediction

Features

• Encoding and decoding from multiple WAV and AIFF formats.
• Support file or stdin/stdout
• Real time processing, encode and decode from streams.

• Configurable

  • LPC order
  • Frame length
  • Number of subframes
  • Adaptive codebook size
  • Multiple frame windows ('hanning', 'boxcar', etc)
  • Weighting coefficients

Dependencies

• numpy
• scipy
• ctlibsnd (ṕython libsndfile wrapper) https://bitbucket.org/san/ctlibsnd

Future improvements

• Implement fixed codebooks of different sizes
• Support more sample rates (currently only 8 kHz)
• Preprocesing filter (high pass, and equalization)

• Convert LPC (linear prediction coefs) to LSP (linear spectral pairs)

  • Quantize LSP
  • Convert LSP to LPC