Classifying Sound Sources Based On Directions Using Audio Visual Stimulus

There are times when a person is in an environment where they must focus on one conversation when multiple others are happening around them. This is referred to as the cocktail party phenomenon. Individuals with impaired hearing lack this ability. This paper gives insight into how the brain handles these situations and how it filters out what a person is not focusing on. Three video monitors were placed in front of the subject and each source played video and audio. The objective was to find any changes in the accuracy of classification when video stimuli are provided. Using a g. Nautilus headset and multiple audio and video sources, EEG is collected from a subject. This data is collected for each data source. Each dataset is used to train a single machine learning classifier which distinguishes the source of sound with a certain accuracy. The results yield an average accuracy of 94.28%