Real-time Coding for Kinesthetic and Tactile Signals

Abstract: The Tactile Internet is at the core of the 5G era, when the world will experience paradigm shift from content-delivery networks to service/labour-delivery ones. Systems that enable wireless communications of haptic data feature bi-directionality, high packet rate and resolution, large degrees of freedom, and above all, strict latency requirements in many applications, aggravating the shortage of wireless resources. Thus, more efficient haptic data reduction techniques are continuously summoned for. Previous studies  on haptic compression mostly resort to DPCM/ADPCM plus entropy coding and perception-based down-sampling for real-time scenarios, and model-based techniques such as DCT and LP for the rest. However, with few exceptions they always segregate tactile signals from kinaesthetic signals, employing only kinaesthetic feedbacks in real-time compression experiments. In addition, these techniques are not optimized for efficient performance at scale. This thesis project proposes a novel multi-channel real-time haptic compression system aimed at teleoperation applications with both kinaesthetic and tactile feedbacks. It consists of a lossy compression layer featuring predictive coding and a lossless layer featuring channel reshuffle and group transmission. By using different quantizer designs in the lossy layer, it abates the need for entropy coding, and leave room for future perception-based data compression modules. The lossless layer exploits inter-channel sparsity for further data reduction. The system is evaluated on a tactile texture database published by University of Pennsylvania in MATLAB. The performance measurements are in both time and frequency domain, mostly objective, but include subjective considerations as well.