Exploit multi-sensory information to improve user experience in
• Speech-centric human computer interaction
• Computer-mediated human inter-communication
- Understand end-users’ requirements
- Identify sensor(s) requirement
- Prototype new hardware
- Develop robust technologies
- Publication 1: Air-and-Bone Conductive Integrated Microphones for Robust Speech Detection and Enhancement in the IEEE Automatic Speech Recognition and Understanding Workshop (ASRU03), November 30 – December 4, 2003.
- Publication 2: Multi-Sensory Microphones for Robust Speech Detection, Enhancement and Recognition in the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2004), May 17-21, 2004.
- Publication 3: Direct Filtering for Air-and-Bone Conductive Microphones in the IEEE International Workshop on Multimedia Signal Processing (MMSP’04), Sep.29-Oct.1, 2004.
- Publication 4: Nonlinear Information Fusion in Multi-Sensor Processing Extracting and Exploiting Hidden Dynamics of Speech Captured by a Bone-Conductive Microphone in the IEEE International Workshop on Multimedia Signal Processing (MMSP’04), Sep.29-Oct.1, 2004.
Sample Waveforms (pending migration)
- Spectral subtraction
- Direct filtering
- Minimum mean square error (MMSE) estimation with a single Gaussian
- Minimum mean square error (MMSE) estimation with a mixture of four Gaussians