Attention Monitor

Summary

To help people improve and optimize their attention, we created a product that utilizes a machine learning model interfaced with music to help train users’ attention. To train the ML model, we used an open-source EEG dataset from Aci et al., (2019) with over 7 million time points and analyzed it using the MNE library based on its frequency and amplitude records. In our analysis, we filtered out the noise and applied the Fourier Series to form our final data in 7 channels. After finishing data analysis, we compared two supervised classification algorithms: support vector machines (SVM) and neural networks (NN). We decided on SVMs due to its reliability and effectiveness in creating accurate models from EEG data. We initially planned on varying the different parameters of scikit-learn’s SVC to determine the best model, such as the type of kernel used (i.e. linear, RBF, and polynomial). Additionally, real-time EEG data was collected using the OpenBCI headset and Ganglion board to determine if the model can accurately predict the level of the user’s attention. Meanwhile, our team developed a headset connection platform and a music interface with Brains@Play to provide feedback.

Built With

• eeg
• mne
• openbci
• pandas
• svm