Turbo detection of space-time trellis-coded constant bit rate vector-quantised videophone system using reversible variable-length codes, convolutional codes and turbo codes

Abstract

Listen

Translate article

We characterise the achievable performance of a proprietary video transmission system which employs a constant bit rate (CBR) video codec that is concatenated with one of three error correction codes, namely a reversible variable-length code (RVLC), a convolutional code (CC) or a convolutional-based turbo code (TC). The CBR video codec is invoked in conjunction with space-time trellis coding (STTC) designed for transmission over a dispersive Rayleigh fading channel. At the receiver, the channel equaliser, the STTC decoder and the RVLC, CC or TC decoder, as appropriate, employ the max-log maximum a-posteriori (MAP) algorithm and their operations are performed in an iterative 'turbo detection' fashion. The systems are designed to maintain similar error-free video reconstruction qualities. which are found to be subjectively pleasing at a peak signal-to-noise ratio (PSNR) of 30.6 dB, at a similar decoding complexity per decoding iteration. These design criteria are achieved by using different transmission rates, with the CC and TC schemes having a 22% higher bandwidth requirement. The results demonstrate that the TC, RVLC and CC systems achieve acceptable subjective reconstructed video quality associated with an average PSNR in excess of 30 dB for E/sub b//N/sub 0/ values above 4.6 dB, 6.4 dB and 7.7 dB, respectively. The design choice between the TC and RVLC systems constitutes a trade-off between the increased error resilience of the TC-based scheme and the reduced bandwidth requirement of the RVLC-based scheme.