{"title":"Fast Factored DCT-LMS Speech Enhancement for Performance Enhancement of Digital Hearing Aid","authors":"Sunitha. S.L., V. Udayashankara","country":null,"institution":"","volume":10,"journal":"International Journal of Biomedical and Biological Engineering","pagesStart":559,"pagesEnd":564,"ISSN":"1307-6892","URL":"https:\/\/publications.waset.org\/pdf\/2325","abstract":"Background noise is particularly damaging to speech\r\nintelligibility for people with hearing loss especially for sensorineural\r\nloss patients. Several investigations on speech intelligibility have\r\ndemonstrated sensorineural loss patients need 5-15 dB higher SNR\r\nthan the normal hearing subjects. This paper describes Discrete\r\nCosine Transform Power Normalized Least Mean Square algorithm\r\nto improve the SNR and to reduce the convergence rate of the LMS\r\nfor Sensory neural loss patients. Since it requires only real arithmetic,\r\nit establishes the faster convergence rate as compare to time domain\r\nLMS and also this transformation improves the eigenvalue\r\ndistribution of the input autocorrelation matrix of the LMS filter.\r\nThe DCT has good ortho-normal, separable, and energy compaction\r\nproperty. Although the DCT does not separate frequencies, it is a\r\npowerful signal decorrelator. It is a real valued function and thus\r\ncan be effectively used in real-time operation. The advantages of\r\nDCT-LMS as compared to standard LMS algorithm are shown via\r\nSNR and eigenvalue ratio computations. . Exploiting the symmetry\r\nof the basis functions, the DCT transform matrix [AN] can be\r\nfactored into a series of \u00b11 butterflies and rotation angles. This\r\nfactorization results in one of the fastest DCT implementation. There\r\nare different ways to obtain factorizations. This work uses the fast\r\nfactored DCT algorithm developed by Chen and company. The\r\ncomputer simulations results show superior convergence\r\ncharacteristics of the proposed algorithm by improving the SNR at\r\nleast 10 dB for input SNR less than and equal to 0 dB, faster\r\nconvergence speed and better time and frequency characteristics.","references":null,"publisher":"World Academy of Science, Engineering and Technology","index":"Open Science Index 10, 2007"}