Knowledge-Based Pitch Detection.
Abstract
Many problems in signal processing involve a mixture of numerical and symbolic knowledge. Examples of problems of this sort include the recognition of speech and the analysis of images. This thesis focuses on the problem of employing a mixture of symbolic and numerical knowledge within a single system, through the development of a system directed at a modified pitch detection problem. For this thesis, the conventional pitch detection problem was modified by providing a phonetic transcript and sex/age information as input to the system, in addition to the acoustic waveform. The Pitch Detector's Assistant (PDA) system that was developed is an interactive facility for evaluating ways of approaching this problem. The PDA system allows the user to interrupt processing at any point, change either input data, derived data, or problem knowledge and continue execution. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1986
- Accession Number
- ADA169069
Entities
People
- Webster P. Dove
Organizations
- Massachusetts Institute of Technology