A neural network algorithm for enhancing delta modulation/LPC tandem connections

The enhancement of coded speech which has been degraded by a tandem connection between wideband and narrowband data links, specifically a delta-modulation (CVSD 16000 b/s) to linear predictive coding (LPC) (2400 b/s) link, is discussed. Speech enhancement is achieved by developing mappings from vector quantization (VQ) codebooks trained on distorted data to VQ codebooks trained on undistorted data. Two different algorithms are adapted to the problem of generating these codebooks. These algorithms are the self-organizing feature map algorithm and learning vector quantization.<
>