Use of Kalman discrete time and frequency filters combined with spectral power subtraction in the reduction of additive noise in speech signals contaminated by colored noise

• Autor(a) principal: Silva, Leandro Aureliano da
• Data de Publicação: 2022
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Research, Society and Development
• Texto Completo: https://rsdjournal.org/index.php/rsd/article/view/30369
• topic: Filtro de Kalman; Subtração espectral; Enriquecimento da voz; Supressão de ruído.; Filtro de Kalman; Sustracción espectral de potencia; Enriquecimiento de voz; Supresión de ruido.; Kalman filter; Spectral power subtraction; Speech enrichment; Noise suppression.

In This article aims to present and compare noise reduction techniques in the reconstruction of speech signals contaminated by colored noise in order to verify the spectral distortion generated by the processed signal. The discrete time and frequency Kalman filters in conjunction with the power spectral subtraction technique were used for noise reduction in signal reconstruction and compared. The signals used were contaminated by colored noise and the performance evaluation of the algorithms was performed using the segmented signal-to-noise ratio and the Itakura-Saito distance that verifies the spectral distortion in the processed signal. The best method found through the tests performed was the discrete time Kalman filter method together with the spectral power subtraction in relation to spectral distortion.