Patent US 7,124,078

A speech coder comprises a first speech enhancement module receiving a digitized speech signal and producing a first enhanced speech signal, a second speech enhancement module receiving the digitized speech signal and producing a second enhanced speech signal, an analysis module that receives the second enhanced speech signal and produces an analyzed speech signal, a first weighting filter that receives the second enhanced speech signal and the analyzed speech signal and produces a second weighted enhanced signal, a quantizer module that receives the analyzed speech signal and produces a quantized speech signal, a synthesizer module that receives the quantized speech signal and an error signal to produce a synthesized speech output signal and a second weighting filter that receives the synthesized speech output signal to produce a first weighted enhanced signal. A subtraction module calculates a difference between the first weighted enhanced signal and the second weighted enhanced signal which difference is utilized to calculate the error signal.