In-Car Speech Separation via MVDR

In this work, a comparative study about in-car microphone array beamforming is carried out. The main objectives of this work are to show how the number of microphones and the array geometry impact in the task of speech separation via beamforming in in-car environment. These studies lead to suggestions about which topologies are more indicated to be applied in real-world automobile applications, aiming the nextgeneration autonomous cars. The beamformer considered is the broadband version of MVDR algorithm and three types of arrays, located at the car console central position, are simulated: linear arrays, circular arrays and elliptic arrays.