METHOD AND APPARATUS FOR COMPENSATING ENERGY FEATURES FOR ROBUST SPEECH RECOGNITION IN NOISE ENVIRONMENT

摘要

Energy Features compensation method and apparatus for robust speech recognition in noisy environments. Characterized in that the energy compensation method comprises: (a) for converting the energy characteristic of the speech training data collected from a noise-free environment as clean environment similar to the energy of the phase noise environment; (B) the minimum value of the speech data by recognizing the energy check is greater than the predetermined minimum value of the target, the energy minimum of the audio data is less than the target minimum, but less than the intermediate value of the energy of the minimum and maximum values of conversion by the conversion equation ERN performing a Half-ERN; And (c) the energy minimum of the speech recognition data, characterized in that it comprises the step of compensating for the energy characteristics not less than the target minimum value by a predetermined method. ; The present invention is more robust than the energy characteristic compensation is possible in a variety of environments it is possible to improve the speech recognition rate in noisy environments.