IVAS: Facilitating Safe and Comfortable Driving with Intelligent Vehicle Audio Systems

Vehicle audio systems affect safety and comfortability of daily driving, as a driver’s mental or physical state can be significantly influenced by his or her audio perception during the maneuver. In this paper, we propose an intelligent vehicle audio systems (IVAS) by leveraging recent advances in automated driving, recommender system, and audio synthesis in an end-to-end fashion. Our contributions are multifold: (i) we design a deep deterministic policy gradients (DDPG) for adaptive control of an automated vehicle equipped with multi-sensors; (ii) we incorporate an unidirectional long short-term memory (LSTM) units of a recurrent neural network (RNN) with a recurrent output layer for audio synthesis; and (iii) we integrate above techniques into an end-to-end framework and pioneer the next generation vehicle audio system—IVAS. A psychological study is also performed to demonstrate the effectiveness of the proposed framework.