Evaluation of visualization algorithms for CommSense system

Application specific instrumentation (ASIN) makes use of sensors and AI (SensAI) algorithms for a highly specialized application, using less computational overhead, it can give good performance. This work evaluates the performance of communication based sensing (CommSense) system using Principal Component Analysis (PCA), kernel PCA (KPCA), t-distributed Stochastic Neighbour Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP) algorithms and their quality of projection. In this paper, we have used Earth Mover's Distance (EMD) (also known as 1st Wasserstein Distance (WD)) for assessing the projections and we reach at the conclusion that, in terms of implementation PCA is the best, but for visualization KPCA, t-SNE and UMAP perform better than PCA.