Application of Spatial Filtering to Speech Recognition.
Abstract
A new approach to the automatic speech recognition problem is presented, based on methods previously successful in visual pattern recognition. Samples of ten different words from seven different speakers were transformed by an electronic analog ear into patterns representing vibratory activity along the basilar membrance of the cochlea as a function of time. The resulting patterns were then Fourier transformed, filtered, compared with a set of stored prototypes, and classified using an adaptive decision function. The resulting recognition rate was evaluated. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1970
- Accession Number
- AD0875082
Entities
People
- David Leroy Thorne
Organizations
- Air Force Institute of Technology