METHOD AND APPARATUS FOR RECOGNIZING WORD SEQUENCE OF AUDIO SIGNAL

摘要

PURPOSE: To effectively reduce searching space together with minimum available look-ahead activity by recognizing a sequence expressed by various 1st strings not canceled based on a related minimum value between 1st scores. CONSTITUTION: A sound signal obtained through a microphone 10 is preprocessed by a device 12, especially a spectral component is digitized in a continuous sound signal section and a test signal is formed by all values expressing respective sound signal sections. The test signal existing in a line 13 is supplied to a device 14 based on look-ahead evaluation and the test signal is supplied also to a recognition device 20 through a delay line 18 for delaying it by the prescribed number K of signals. A word sequence recognized by the device 20 is supplied to an output device 22, e.g. a display device or a printer. Thus the sequence expressed by various noncanceled 1st strings is recognized.