A Current Prediction Model Based on LSTM and Ensemble Learning for Remote Palpation

As an important technology of virtual reality, tactile reproduction enables users to touch and perceive virtual objects. The emergence of remote palpation technology based on tactile reproduction provides a new idea for disease diagnosis. Remote palpation technology collects tactile perception from the patient side, transmits it to the doctor side through the network, and uses various tactile devices to reproduce it. However tactile reproduction has high accuracy and real-time requirements. Our group designed a handheld tactile perception and reproduction system for remote palpation. This system generates tactile force feedback by driving coil current in an electromagnetic tactile device. Therefore, it is very important to establish a fast and accurate current prediction model to generate tactile feedback. We observe temporal relationships in tactile information, so we take advantage of the time series prediction model LSTM and the regression prediction model GRNN, and use the idea of ensemble learning to build a more powerful and accurate current prediction model. We conduct comprehensive experiments, and the experimental results show that our proposed method helps to improve the accuracy and speed of remote palpation.