We are currentcly considering creating a speech database for speech recognition in meeting room environment. What we have at hand is a high-quality database collected by close-talk microphone. What we going to do is to playback the existing speech files via a speaker (or speakers in different locations ) and re-collected by a far-talk microphone (or microphone array). Are there any suggestions for speaker and microphone products for this project? We hope that the ouput of the speaker could be as close as possible to the natural speech, which means that we may need to equalize the frequency response of the close-talk microphone and the speaker. To your experience, is it possible to achieve this goal by equalization or any ohter signal processing techniques?
Digital Signal Processing & Speech Technology Laboratory
Department of Electronic Engineering
The Chinese University of Hong Kong
Shatin, NT, Hong Kong SAR of China