We are creating a fairly sophisticated transcription program. Key
characteristics of the data set include:
- Two speakers are being transcribed, each with their own
microphone, with significant crosstalk between microphones;
- A video file of the interaction, and
- A stroke file from wacom tablet data that's being captured.
We have the basic transcriber running very well.
We now want to extend the program so we can pro grammatically process
the audio data:
- automatically decide who is speaking
- skip over silence in the audio stream
- display the audio waveform for each of the audio channels in the
audio track.
My question: how do I get access to the audio sample data? I would
like two access modes:
- realtime, where we're displaying the waveforms as the movie is
being played (where if we reposition playback, I need to get 10 seconds
of data before and after the new start point), and
- offline, where we want to quickly process all of the audio
stream, extracting characteristics. In this case, we don't want to
spend two hours playing through all of the audio data.
So, how do I
- get the samples, and
- how do I know how to interpret them?
Thanks in advance; we greatly appreciate any help,
John Newkirk
|
