In the span of the BCI Hackathon, we aimed to experiment with different dimensionality reduction techniques, in order to explore possible interpretation of the Willett Speech Dataset from geometrical and neurological perspective innovatively. The Willett Speech Dataset[1] contains Phenomes data, Orofacial data and FiftyWord data and we chose to focus on the Phenomes data, which studies participants attempted to speak single phonemes in response to cues shown on a computer monitor. Our main analysis was performed on Spike power band (termed "spikePow" in the dataset) and aimed to apply topological data analysis techniques on the neural signals to reduce the 40 pairs of dimension in 16 * 256 * 100 to a lower dimension.