Neural Network Implementation

Using Python as our programming language of choice because it offers the highest versatility, we built a neural network using Keras (specifically TensorFlow) and sample data generated from our team. First, we took the Fourier transforms of the data, allowing us to more easily identify important signal components. Next, the data was fed into 3 independent linear classifiers (one for each emotional coordinate). The model then output a vector containing scalar values for each of the emotional coordinates. This vector was directly input into TouchDesigner (an animation software) to generate the final visual output.