Alaryngeal Speech Enhancement Using Minimum Statistics Approach to Spectral Subtraction
Alaryngeal Speech Enhancement Using Minimum Statistics Approach to Spectral Subtraction
Authors: Hamed Azarnoush, Faraz Mir, Sos Agaian, Mo Jamshidi, Mehdi Shadaram
Description: The objective of this paper is to propose an approach for enhancement of speech generated by an ALT (Artificial Larynx Transducer). Spectral subtraction using Martin’s statistical method is chosen as the means to confront the speech enhancement problem. In contrast to the conventional spectral subtraction methods, Martin’s approach does not need a VAD (voice activity detector) nor histograms to learn statistics. Since the major power of noise is present on the voiced segments of the signal, the conventional spectral subtraction approaches might face difficulty dealing with this problem. Therefore, Martin’s algorithm is particularly powerful in noise rejection from ALT-produced signal. Successful implementation of alaryngeal speech enhancement using Martin’s approach is presented through some simulations.