DCT based Algorithm on Dimension Reduction of LSF Parameters

A dimension reduction method of LSF parameters based on DCT is proposed to decrease codebook storage and search complexity of LSF.Compared to the existing method of taking out and transmitting important frames in time domain,DCT based algorithm can decrease codebook storage and search complexity by 50%,without significant decrease of hearing performance.Tests show that average spectral distortion is reduced by about 0.1dB to 0.2dB by using the algorithm in this paper instead of using method of taking out and transmitting important frames in time domain.The results verify that the DCT based algorithm can efficiently decrease codebook storage and search complexity of LSF parameters in speech codec.