Key Word Classification.

Abstract

This report summarizes the development of a real-time key word recognition. The basic objectives for the system were the ability to detect one or more key words in continuous speech, independent of the speaker. By concentrating half-syllable sized utterances in sequence to detect the occurrence of a spoken work, good temporal registration was obtained and with good recognition results. On a limited number of talkers (about ten) results of 90% - 95% detection and 4-6 false alarms per hour were obtained. When the data base is increased to approximately fifty talkers, the results obtained are 85% detection and 20-25 false alarms per hour. The techniques and hardware developed under this effort appear very promising. Improvements appear feasible with only modest changes. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 01, 1977
Accession Number
ADA044605

Entities

People

  • P. N. Leiby
  • R. E. Smith
  • S. L. Moshier

Tags

Communities of Interest

  • C4I
  • Human Systems
  • Space

DTIC Thesaurus Topics

  • Accuracy
  • Arithmetic Units
  • Arrays
  • Artificial Intelligence
  • Assembly Languages
  • Computer Programming
  • Computer Programs
  • Computers
  • Confidence Limits
  • Data Science
  • Databases
  • Detection
  • Distribution Functions
  • False Alarms
  • Information Science
  • Operating Systems
  • Pattern Recognition

Fields of Study

  • Biology

Readers

  • Computational Linguistics
  • Sensor Fusion and Tracking Systems.
  • Systems Analysis and Design