SpeakerBeam: A New Deep Learning Technology for Extracting Speech of a Target Speaker Based on the Speaker's Voice Characteristics

In a noisy environment such as a cocktail party, humans can focus on listening to a desired speaker, anability known as selective hearing. Current approaches developed to realize computational selectivehearing require knowing the position of the target speaker, which limits their practical usage. This articleintroduces SpeakerBeam, a deep learning based approach for computational selective hearing based onthe characteristics of the target speakers voice. SpeakerBeam requires only a small amount of speechdata from the target speaker to compute his/her voice characteristics. It can then extract the speech ofthat speaker regardless of his/her position or the number of speakers talking in the background.