Haptic Data Compression Based On Quadratic Curve Reconstruction

Haptic can provide the contact perception information of virtual objects. By using haptic interface, users are not only able to see the remote or microscopic objects through the screen, but also able to touch and manipulate them, which enhances the users’realism and immersive more directly. The haptic data signal contains multidimensional information, such as the position, angle, velocity, angular velocity, acceleration, angular acceleration, the feedback force and torque data within the passing of time, and so on. The goal of haptic performance is to make the user get a realistic tactile sensation in the contact interaction with virtual objects. Because of the tactile sensitivity of human, haptic rendering needs at least1kHz update frequency to give people sense of reality. However, in the remote control system with bidirectional data transmission, the remote digital communication network is difficult to maintain such a high packet rate (1000packets/sec), and unlike soft real-time audio and video streaming, haptic data streams belong to the hard real-time, which means even minimum delay could lead to the instability of telepresence and teleaction (TPTA) system. Therefore, the characters of multidimensional, high frequency and data-intensive make it an urgent need to compress online haptic data, through which we can reduce haptic data traffic, improve the performance of remote haptic interaction system and maintain a high quality user experience in a TPTA system.Haptic data compression is a fairly new research area that has attracted much interest in recent years. Haptic data compression includes online and offline compression. We focus on the offline haptic interaction data compression. The compression of haptic data is still in its infancy. The wide-used haptic data compression algorithm is based on deadband principle and linear prediction model, while the haptic data is highly nonlinear, and consequently, using a basic linear prediction model would generally not yield optimal results. Based on deadband, the downsampled signals leads to a loss of information and distortions, which makes discontinuities and sudden changes of the upsampled signals by the reconstruction model.This thesis aims to solve the above problems and presents two haptic data compression methods.Method Ⅰ:A prediction model based on quadratic curve is constructed, based on the prediction model together with the perception deadband, the proposed method can compress the haptic data, and the experimental results illustrate the efficiency of the proposed method.Method Ⅱ:We propose a haptic data compression based on quadratic curve reconstruction with C0continuity in this paper. The algorithm first divides the haptic data into subsets based on the continuous feature of the actual signal, then divides each continuous subset into smaller subsets based on the prediction model of method Ⅰ. In a small subset, we could predict all haptic data based on the same predictor, and each continuous subset is fitted by a C0piecewise quadratic curve. Unlike the common compression methods which store the actual sample data, our method store the coefficients of the piecewise quadratic curve in order to compress the data; the proposed algorithm takes full account of continuity between data sets and produces a smaller signal distortion after decompression, the computing instances illustrate the efficiency of our method.