Biologically Based Non-Language Speech Sound Detection

This proposal addresses the application of new acoustic processing technologies to automatically identify and eliminate non-language speech sounds as a pre-processing stage to improve audio processing. Non-language speech sounds make up a large part of natural human language use, but contemporary speech recognition data preparation relies on hand-labeling of non-language speech sounds. The proposed work will extend the capabilities of a computational model of auditory processing based on multiscale spectro-temporal modulation features to the automated detection of non-language speech sounds. Advanced Acoustic Concepts and the University of Maryland have extensive experience applying the computational model to a variety of sound recognition and sound stream segregation problems. One particularly relevant application involved distinguishing speech from non-speech sounds consisting of animal vocalizations, music, and environmental sounds. Using a sound database to be provided by the sponsor, algorithms will be trained and tested in a similar fashion to distinguish non-language speech sounds from language. Algorithms for exact segmentation of the sound stream will be designed and demonstrated.