11-756 / 18799D Design and Implementation of ASR Systems

11-756/18799D ASR: Assignment 0, data capture

Problem

Write a routine to perform live data capture.

The data must be captured at 16khz sampling rate
It must be 16-bit PCM data
Recording must be "hit-to-talk"
- The user strikes a key to begin recording
- The termination of the recording must be automatically determined.

The output must be stored in a file.

Some suggestions

The initial recording immediately after a keyboard hit may have a keyclick. Deleting the first 5-10ms of recording (more, depending on your keyboard) may get rid of it.

Use the adaptive endpointing algorithm in the lecture slides. You may have to tune update and forgetting factors to get it to work right.

Visualizing the recorded data always helps -- problems in endpointing will immediately become apparent.

You can use portaudio for data capture.

Due: Wednesday, 8 Feb 2011.