Haptic Data Compression Technique Based On A Linear Prediction Model And Quadratic Curve Reconstruction

In today’s multimedia systems, visual and auditory information are predominant. Efficient data compression methods and standards have paved the way for widespread adoption of digital audio and video. However, there’re not only visual and auditory in real life, haptic also plays a very important role. Relative to the study in full swing of the field of audio and video, the field of haptic interaction study is few interested. As the latest technology in the field of human-computer interaction, haptic interaction studies how to enhance the exchange of people with computers and robots by using tactile information, in applications such as surgical simulation, telemedicine, entertainment, robot and job training.This paper mainly presents the compression of offline haptic interaction data. The haptic data we deal with is a kind of high-dimensional discrete mass data cloud, which sampled from the continuous haptic interaction data. The basic signal of haptic data contains position, angle, feedback force and torque data within the passing of time. Different from the current widespread use of simple linear prediction, we construct a pre-conic tangent liner model, which is better reflects the trend of the actual data. Based on this model, the hapitc data are divided into several subsets. By re-fitting each subset of data makes forecast data closer to the original data, and recovery better. Meanwhile, by using geometric distance instead of Euclidean distance strengthens the compression ratio.The experimental results show that, compared with LSE-LP method, our approach gains greater compression ratio, smaller relative error and better recovery.