Control Of Robotic Arm Based On Steady-State Visual Evoked Potential

Brain-Computer Interface technology provides a new way for people with disabilities to interact with the surrounding environment.Steady-State Visual Evoked Potential has been increasingly applied to BCI systems due to its distinct characteristics,low data requirements,and high classification accuracy.However,the current SSVEP-based BCI systems suffer from long gaze-stimulus interface time,complex GUI interface design,and low task execution efficiency.To provide better user experience for disabled individuals,this thesis presents a SSVEP-based robotic arm grasping system,with the following main contributions:During the process of SSVEP signal acquisition,factors such as environment,equipment,and manual operation can have a significant impact,making the stability of the model particularly important for different signal-to-noise ratios.This paper proposes a Spatial Filter Bank Canonical Correlation Analysis model,which extracts four types of spatial filters from the original signal,artificial synthesized signal,and averaged feature signal for subsequent classification.The model is compared with four other models under different signal-to-noise ratios.Six low signal-to-noise ratio datasets were self-collected and six high signal-to-noise ratio datasets were obtained from the Tsinghua University dataset.The experimental results show that the proposed model has excellent classification performance when using high and low signal-to-noise ratio data,with the highest classification accuracy reaching 99.24%.The model also demonstrates high robustness to different data lengths,and the maximum information transfer rate(ITR)can reach 105.1 bits/min.Subsequently,the best weight coefficient combination is obtained through grid search to improve the accuracy of the model by 10% when the data length is 2 seconds.To reduce the gaze time and operational difficulty of subjects,a robotic arm system was designed for automatic target grasping.This system consists of two parts: motion and vision.The vision system is responsible for calculating the world coordinates and rotation angles of the target grasp point,while the motion system solves the coordinate parameters of the robotic arm and controls the movement of each joint.To improve the success rate of grasping tasks,a designed motion strategy is followed by the robotic arm during the execution of the grasping task.The experiments were conducted under ideal and poor lighting conditions,and the grasping accuracy reached 99.3% and 84.7%,respectively,indicating that an ideal lighting environment is conducive to the execution of the grasping task.Using the classification results of SSVEP signals to control the movement of a robotic arm.The accuracy of each stimulus frequency of the self-collected data was calculated,and the results indicated that the data at 13 and 17 Hz were not suitable for subsequent experiments.An offline experiment was conducted to simulate the system’s operation and the results of the self-collected dataset and the Tsinghua dataset were 88.8% and 94.4%,respectively.In the online experiment,the GUI only used four stimulation matrices corresponding to three colors of targets and ending grasping.Under the condition of a short stimulus matrix and viewing time,the subject was able to complete the grasping task relatively easily.The final success rate of accurately identifying the color,completing the grasping and placing task was 83%,with an average grasping time of 26.4 s,indicating that the system has excellent performance and certain practical value in actual applications.