Smart Wheelchairs And Brain Computer Interfaces PDF Download

Smart wheelchairs with semi or fully autonomous functions, can greatly improve the mobility of persons with physical impairments. However, most are controlled using inputs that require physical manipulation (e.g. joystick controllers) and for persons with severe physical impairments this method of control can be too demanding. A noninvasive brain-computer interface (BCI) technology-based controller could bridge the gap between smart wheelchairs and physically impaired persons with severe conditions. Current BCI-controlled wheelchairs rely on detecting steady-state visually evoked potential (SSVEP) responses as these typically have the greatest data transfer rate. However, this method requires the user to focus on a screen for an extended period of time. This causes strain on the user and takes their attention away from their surroundings, which could be dangerous in a scenario that requires navigation around multiple moving objects. The main goal of this project is to design a hybrid BCI controller using an electroencephalogram (EEG) headset to detect hand motor imagery (MI) and jaw electromyography (EMG) signals to control a smart wheelchair in conjunction with its semi-autonomous capabilities. A controller of this kind is well-known to have low data transfer rates, and therefore has lower accuracy and longer response times as compared to other controllers. However, a properly structured controller hierarchy between the BCI controller and semi-autonomous system is developed to compensate the limitations of the controller's accuracy. OpenViBE, an open source BCI programming application, and a commercial grade EEG headset are used for data acquisition. A multiple common spatial pattern (CSP) filter and linear discriminant analysis (LDA) classifier system are chosen to process and classify the user's brain activity. Users are trained to use the system and classifier profiles are optimized for each user. Two subjects are trained to use the BCI system and the results of their practice are analyzed for this study. The average classifier accuracy for Subject 1 is 0.87 and the average classifier accuracy for Subject 2 is 0.83. Next, a fuzzy logic controller (FLC) controller is created in LabVIEW to convert the classifier data to a signal that is compatible with the semi-autonomous wheelchair system. The performance of the controller scheme is evaluated using the OpenViBE "Replay" script and recorded training data. Positive, negative, and false positive executions are recorded for each subject. Initially, positive rates for both subjects were strong, but false positive rates were too high to be used. The design is iterated by changing the configuration of the LabVIEW script and the rules of the FLC. The configuration with the best positive rates for turn executions is chosen. The average positive rate for turning is 0.68 for Subject 1 and 0.64 for Subject 2. The positive rate for executing the start/stop function, however, was only able to achieve 0.50 for Subject 1 and 0.38 for Subject 2. False positive rates, however, were brought down to levels below 0.20 for both subjects.