5.2 Linear prediction analysis and quantization

26.1903GPPAdaptive Multi-Rate - Wideband (AMR-WB) speech codecSpeech codec speech processing functionsTranscoding functionsTS

Short-term prediction, or LP, analysis is performed once per speech frame using the autocorrelation approach with 30 ms asymmetric windows. An overhead of 5 ms is used in the autocorrelation computation. The frame structure is depicted below.

The autocorrelations of windowed speech are converted to the LP coefficients using the Levinson-Durbin algorithm. Then the LP coefficients are transformed to the ISP domain for quantization and interpolation purposes. The interpolated quantized and unquantized filters are converted back to the LP filter coefficients (to construct the synthesis and weighting filters at each subframe).