Topology-enhanced machine learning for consonant recognition

In artificial-intelligence-aided signal processing, existing deep learning models often exhibit a black-box structure. The integration of topological methods serves a dual purpose of making models more interpretable as well as extracting structural information from time-dependent data for smarter learning. Here, we provide a transparent and broadly applicable methodology, TopCap, to capture topological features inherent in time series for machine learning. Rooted in high-dimensional ambient spaces, TopCap is capable of capturing features rarely detected in datasets with low intrinsic dimensionality. Compared to prior approaches, we obtain descriptors which probe finer information such as the vibration of a time series. This information is then vectorised and fed to multiple machine learning algorithms. Notably, in classifying voiced and voiceless consonants, TopCap achieves an accuracy exceeding 96%, significantly outperforming traditional convolutional neural networks in both accuracy and efficiency, and is geared towards designing topologically enhanced convolutional layers for deep learning speech and audio signals.