Patent №
US 4,028,496
Granted
—
Owner
—
Lab
—
AI components
1
speech
Assignment
None on record
Dataset
AIPD
2023_r1 edition
Application
05715121
The detection sensitivity and noise rejection of an arrangement for detecting speech in the presence of noise is improved by accumulating the weighted differences between input signal samples and their short-term running average. The detector thus tracks ambient noise, providing an adaptive detection threshold such that detection sensitivity is increased in low noise environments without excessive false operation on high level noise. The peak average attained during an interval of speech is used to provide variable hangover upon cessation of speech, yielding greater hangover for weak talkers than for loud talkers. In an illustrative embodiment of the speech detector used in a speech interpolation system, protection is afforded also against false transmission path operation due to detection of speech echo.