This paper outlines preliminary steps towards the development of an audio-based room-occupancy analysis model. Our approach borrows from speechrecognition tradition and is based on Gaussian Mixtures and Hidden MarkovModels. We analyze possible challenges encountered in the development of such amodel, and offer several solutions including feature design and predictionstrategies. We provide results obtained from experiments with audio data from aretail store in Palo Alto, California. Model assessment is done vialeave-two-out Bootstrap and model convergence achieves good accuracy, thusrepresenting a contribution to multimodal people counting algorithms.
展开▼