Browse Prior Art Database

Spoken Word Recognizer

IP.com Disclosure Number: IPCOM000073423D
Original Publication Date: 1970-Dec-01
Included in the Prior Art Database: 2005-Feb-22
Document File: 2 page(s) / 47K

Publishing Venue

IBM

Related People

Clapper, GL: AUTHOR

Abstract

In a feasible spoken word recognizer, there are two requirements which must be satisfied. First, it is necessary that the output be 100% correct even through there is a probability of some misrecognition due to operator variability and noise and, second, it should be capable of complete hands-off operation during data gathering by voice recognition.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 2

Spoken Word Recognizer

In a feasible spoken word recognizer, there are two requirements which must be satisfied. First, it is necessary that the output be 100% correct even through there is a probability of some misrecognition due to operator variability and noise and, second, it should be capable of complete hands-off operation during data gathering by voice recognition.

Visual verification before word code release and the ability to erase mistakes by voice command can be combined to assure correct operation. The draw shows a block diagram of such a system. Initially, the operators manual controls 13 to select and adapt electronic templates to particular voice or language and for a short vocabulary training can be completed in a few minutes. When the system is sufficiently trained, operation can be continued under voice control alone.

Words are spoken into a microphone 1 and are analyzed for frequency patterns in analyzer 2. The pattern is stored temporarily in storage 3 until the end of the word is detected by a nonlinear time base 4. Time base 4 is started by the detection of the start of a word in analyzer 2 and controls both the sensitivity of the analyzer 2 and the sampling rate of storage 3. When the word is recognized as complete, its stored pattern is compared with the templates stored during the training period in a memory 5. The best match of the word to the stored templates is sent to a symbol store where it is translated into display control characters. A display control 7 receives the decoded character and will present it on a display 8. The...