An improved Mel-scale based multi-band spectral subtraction using formant intensification

Speech enhancement is gaining popularity and has become a prominent exercise because of wide usage of speech in today's communication which is prone to corruption by additive noise picked from surroundings of the speaker. This paper proposes unique Mel-scale based multiband spectral subtraction speech enhancement system using formant intensification and recursive noise estimation in colored noise conditions. The enhancement performance of this particular method is examined using spectrograms, informal hearing tests, objective measures and vows to offer better clean speech estimation than multi-band spectral subtraction technique in vogue. The objective measures obtained through experimental simulation project the improvement and are also supported by informal listening tests.