We build machine-learning tools to correct optical character recognition (OCR) errors and automatically tagg large collections of historical documentsRead More
We investigate techniques to alter foreign-accented utterances to sound more “native” while preserving the unique vocal properties of the speaker. This requires altering both prosodic and segmental characteristics of the speech signal, and perceptual evaluation with human listeners.Read More
This short video describes the ongoing research work in the lab.Watch the clip
We are developing active sensing strategies for tunable chemical sensors, that is, sensors whose selectivity towards different chemical species can be fine-tuned programmatically; this includes metal-oxide chemical sensors under temperature modulation and Fabry-Perot infrared interferometers.Read More
We develop interface circuits, interfaces for mobile platforms, biosignal processing algorithms, pattern recognizers and biofeedback games. We integrate these systems and evaluate them through human studies.
We develop accent-conversion methods for non-native speech, speech therapy tools for children with motor disabilities, and techniques for animating 3D facial models from speech.
We integrate chemosensor systems, develop pattern recognition methods to extract information from sensor signals, and machine learning methods to optimize sensor tunings on-the-fly to adapt to the environment.