Performance comparison of intraframe and interframe LSF quantization in packet networks

Line spectrum frequencies (LSF) have been the prevailing parameter set to represent LPC coefficients in speech coding. Extensive research has been performed to exploit their interframe and intraframe correlations and quantize them more efficiently. Interframe coding of the LSF can cause error propagation when frame erasures occur. Since most LSF quantizers were designed with the primary concerns of bit-rate and complexity, less attention was paid to error propagation. We investigate the erasure performance of interframe LSF coding and compare it with an intraframe coding method. Our results show that with only 5% extra bit-rate, intraframe coding is much more robust to frame erasures and a typical improvement of 0.5 dB on spectral distortion can be obtained with 20% packet loss. Subjective listening tests indicate significant improvement as well.