A Scalable Wideband “Add-On” for the G.729 Speech Codec

Inhalt:
We present a technique to enhance the perceived speech quality in narrowband (NB, cutoff frequency fc = 3.4 kHz) speech communication systems, e.g., implementing the ITU-T G.729A codec. Therefore, wideband (WB, fc = 7 kHz) capability is introduced. The limited acoustic bandwidth of NB speech coding is extended using a fairly coarse description of the temporal and spectral envelopes of the missing high frequency band (3.4–7 kHz). These parameters are quantized, transmitted and then used at the receiver side to synthesize the high frequency components based on a synthetic “excitation signal”. The parameter quantization is done using split Vector Quantization (VQ) in a transformed domain. The quantization scheme can be tailored to match the desired bitrate. We have implemented an example configuration and performed a listening test to compare the approach with common WB speech codecs.