Categories


Project: Speed-Reader

Speed-Reader is an audio book reader that gives the user control of the speaker rate when listening to audio books.  Speed-Reader uses digital speech signal processing to accomplish low distortion timescale modification of speech in real time.  Timescale modification of speech is accomplished by selectively repeating segments of speech to slow down the speaker rate or deleting selected speech segments to speed up the speaker rate.  Digital signal processing is used to determine which speech segments are candidates for repetition or deletion.

License: Apache License, Version 2.0

Hardware: Beagle Board, Which is based on the TI OMAP-3530 applications processor.  The ARM CORTEX is used for I/O and user interface while the C64++ DSP is used for the digital audio processing.

Plan:

  • The speech processing algorithms are currently being protyped using GNU Octave, an open source math tool similar to MATLAB.
  • A C++ version of the algorithms will be developed on a standard Linux desktop in a non-realtime mode writing output to files.
  • Once the algorithms are completed, they will be ported to the Beagleboard DSP.
  • Initially the digitized speech from audio books wll be read using the SD card interface.
  • The output of the speech processing integrated with the Beagleboard audio.
  • User interface will be developed for realtime control of the playback rate.
  • Phase 2 will add a CDROM reader using the USB port to play audio books directly from CDs.