I build systems that help people learn, make decisions, and communicate.
Most of my speech-related work has been done in a research context, and has often involved the CMU Sphinx 4 recognizer.
Projects have included digital signal processor (DSP) platforms, detection of non-lexical speech sounds (NLSS), detection of out-of-vocabulary (OOV) words, adding eye gaze information, and using formal semantic reasoning to reduce the recognition search space.
Some of my code has found its way to GitHub, including helper data for Sphinx 4 acoustic model adaptation, a utility class for word error rate calculation, and code to gather simple speech data containing relative spatial and property semantics.
ItemCounter is a special-purpose Java class designed to make counting objects easy. In addition to accessor methods, it offers min-count, max, mean, count-of-count, and pretty print capabilities.