VOICE RECOGNITION ACCURACY DEGRADATION FACTOR ESTIMATION DEVICE, VOICE RECOGNITION ACCURACY DEGRADATION FACTOR ESTIMATION METHOD, AND PROGRAM

摘要

To provide a voice recognition accuracy degradation factor estimation device capable of estimating a voice recognition accuracy degradation factor from feature of the voice.SOLUTION: The voice recognition accuracy degradation factor estimation device includes a voice feature quantity extracting unit which extracts a voice feature quantity for each frame from the input voice, a posterior probability calculation unit which calculates a posterior probability for each acoustic event of a feature amount for each frame, on the basis of a plurality of acoustic events classified in advance as either a degradation factor class or a non-degradation factor class, a filtering unit to obtain a modified posterior probability by performing filtering by a time series filter which is a filter in which a weighting coefficient is expanded to the posterior probability for each acoustic event in a temporal direction, and a degradation factor output unit which, when an acoustic event is consecutively identical within a predetermined time section in which the modified posterior probability is maximum in each frame in acoustic events classified in degradation factor class, outputs the acoustic event as a degradation factor.SELECTED DRAWING: Figure 1