Multi-sensor weighted support vector machine algorithm oriented to P300 signals

The P300 signal is widely used in brain computer interfaces (BCIs) because of its high recognition accuracy, flexible number of commands and short training time. Mapping P300 signals into control commands, namely, P300 signal processing is the research core of BCIs. Focusing on variability of raw data collected from different electrodes, a multi-sensor weighted support vector machine (msw-SVM) algorithm is proposed. It makes the amplitude difference of targets and non-targets signals more obvious to obtain better recognition accuracy. Experiments proved the classification result of this proposed method outperforms the traditional support vector machine (SVM) method. Meanwhile, as for P300 signal pre-processing, an optimal weighted averaging filter was employed to enhance the signal-to-noise ratio. It offers better data sources for signal processing.