Wide band sub-band speech coding using nonlinear prediction

TL;DR

This paper compares linear and nonlinear neural network-based prediction methods for wide band sub-band speech coding, demonstrating that nonlinear prediction improves signal quality by up to 2 dB SEGSNR and proposing a method to synthesize wide band signals from narrow band inputs.

Contribution

It introduces a nonlinear neural network prediction scheme for sub-band speech coding and a simple method for wide band signal synthesis from narrow band signals.

Findings

Nonlinear prediction outperforms linear prediction by up to 2 dB SEGSNR.

The proposed synthesis method effectively generates wide band signals from narrow band inputs.

Abstract

We compare a wide band sub-band speech coder using ADPCM schemes with linear prediction against the same scheme with nonlinear prediction based on multi-layer perceptrons. Exhaustive results are presented in each band, and the full signal. Our proposed scheme with non-linear neural net prediction outperforms the linear scheme up to 2 dB in SEGSNR. In addition, we propose a simple method based on a non-linearity in order to obtain a synthetic wide band signal from a narrow band signal.