Search audio file for pattern?

In theory—you’ll have to write your own app to do this—you would need to do this:

1. Run the WAV file through a speech recognition engine Sphinx Speech Recognition is a free open source engine.

2. Have your speech recognition engine spit out text in real-time in memory and also keep track of the sample length location. Everytime the string “deux radio” is found, spit out the sample location to a splitlog.txt file. Since this is in realtime you get an approximation of the sample location.

3. Write your own wav splitter app: Gather all the sample location points from splitlog.txt and split the WAV accordingly.

There is a program called audiogrep which transcribes text using speech-to-text conversion and then creates audiocuts based on text searches.

I haven't seen any option to output time stamps instead of audio clips, you probably need to modify the source slightly to that (or contact the author and ask him). I also don't know how well it will handle French (or whatever language your announcer uses).

You could try :

• TAPESTREA: Experimental software which is described as:

  TAPESTREA (or taps) is a unified framework for interactively analyzing, transforming and synthesizing complex sounds. Given one or more recordings, it provides well-defined means to:

  • identify points of interest in the sound and extract them into reusable - templates
  • transform sound components independently of the background and/or other events
  • continually resynthesize the background texture in a perceptually convincing manner
  • controllably place event templates over backgrounds, using a novel - graphical user interface and/or scripts written in the ChucK audio programming language

• Voice Converter: Shareware ($29.95, trial version with no time limit) which is described as a tool that can be a:

  Pitch Editor, Formant Shift, Vibrato, Raspiness, Breathiness, Pitch Range Scaling, Pitch Smoothing, Time Stretching…