Design and Implementation of Normalized Hybrid Projection Algorithm for Speech Enhancement

Ch.D. Umasankar and Dr.M. Satya Sai Ram

The speech enhancement aims to improve the intelligibility and quality of the speech. By recording the speech signal in the noisy environment, clean speech signals degradation is bound to happen. Speech enhancement reduces the noise without distorting the original (clean) signal. In this concept, Adaptive Hybrid algorithm method is proposed to enhance the noisy speech signal. Filtering techniques improvising the predictive data analysis would fundamentally apply on the Noise (Unwanted signal), to enhance specific design criteria on the algorithmic model for a specific filter as FIR we have considered LMS, NLMS, AFFINE PROJECTION algorithms. To analyze and impart such criteria on the problems associated with FIR filtering techniques we have focused on two algorithms (NLMS and AP) which would provide a hybrid algorithm analyzing the faults in NLMS and AP. We propose such changes which are affected only on LMS and APS algorithm. In this design model we have structuralized specific parametric criteria where each scenario of design would improvise the noises ranging from 0, 5, 10, 15 dB utilized in Babble, Factory, Destroy Engine, Car, Fire Engine and Train Noises based on the TIMIT and NOIZEUS Database for each clean signal observed. Based on this efficient architecture for the implementation of a NHP adaptive algorithm have been developed and compared with the existing structure. These works propose a strategy for optimized balanced pipelining across the time-consuming combinational of the fabric on FIR filter. The performance of this algorithm will be compared by considering in terms of SNR, MSE, RMSE, and Distortion of these algorithms. Based on the performance evaluation, the proposed algorithm was found to be a better optimal noise cancellation technique for speech signals.

Volume 12 | Issue 1

Pages: 223-232

DOI: 10.5373/JARDCS/V12I1/20201033