A STUDY OF MEASURES FOR NOISE REDUCTION TO IMPROVE THE QUALITY OF SPEECH USING ADJUSTMENT KALMAN FILTER ác giả , phapdn@gmail com óm tắt i á t t g g i c t t giảm i c t i ti g ói g c m i c i c g m i i[.]
A STUDY OF MEASURES FOR NOISE REDUCTION TO IMPROVE THE QUALITY OF SPEECH USING ADJUSTMENT KALMAN FILTER ác giả: , phapdn@gmail.com óm tắt: i t i c t i g i t i m g i m c i t c ó t g g i i c ti g ói i c g m i c ct c i i i m ảc ừk ó : c t t giảm i c gm i im t i mt ic t cc t c i ác ác t i i ti t c i c i ác g t tt m i t g i ó t i ti g ói g c g ct c i ti g ói c g g g ói t tt mm g t t ti c giá ti i g ó c m c i c tác c g K t ả g i c g i c Abstract: The article focuses on noise reduction techniques for speech using the adjusment Kalman filter The construction of the regression model with matrix of estimated coefficients are made for both speech and noise to change the structure of the filter Original speech affected by noise is processed prior to the input filter using the signal estimation method including the determination of the speech presence probability (SPP) and spectral subtraction technique The simulation results are performed on the NOIZEUS database to simulate the studied technique with different kinds of noise and different noise levels Through the evaluation criteria, the article compares the technical efficiency on noise environments and the methods studied previously Key words: Kalman filter; Noise reduction; Speech; Speech presence probability; Regression