Source enhanced linear prediction of speech incorporating simultaneously masked spectral weighting

Abstract

Linear prediction is the cornerstone of most modern speech compression algorithms. This paper proposes modifying the calculation of the linear predictor coefficients to incorporate a~weighting function based on the simultaneous masking property of the ear. The resultant prediction filter better models the perceptual characteristics of the source and results in the removal of more perceptually important information from the input speech signal than a standard LP filter. When employed in a low rate speech codec the net effect is an improvement in subjective quality, with no increase in transmission rate and only a modest increase in computational complexity.