We present the results of a final project that was intended to evaluate a speaker position inside a vehicle, using semi-supervised learning algorithm.
The dataset that was used in order to create and validate the proposed solution was labeled recordings which were sampled by an array of microphones positioned at the front of the vehicle. The recordings were labeled by different parameters (such as head angle of the speaker, presence of another person in the vehicle, etc). while the main parameter was the speaker position.
The main goal of the project was achieved by using pre-process and data analysis techniques which were examined throughout the project in order to improve the speaker position prediction, while the final result was achieved by classifier learning process based on the processed data.
During the project, the efficiency and contribution of pre-process methods were examined such as the Diffusion Map dimension reduction method and using in Relative Transfer Functions. In addition, the robustness of the classifier to noise was examined as well.