Ihr Startpunkt für Innovation

Generative Deep Learning Models for Signal Processing

Houssam Razouk CTR / FH-Kärnten

Abstract: Using generative deep learning models for signal processing of sound waves is the focus of the talk . A new network design is developed to learn the transfer function of a loudspeaker. The transfer function of loudspeakers is learned by a deep generative network using training pairs from the ground truth data and the recorded (simulated) data from the output of the loudspeaker. This network is used to train a second generative network to learn the inverse transfer function of the same loudspeaker and is used to pre-distort the sound wave signal in order to get the optimal performance from the loudspeaker. Caffe-framework with python interface is used to develop and implement the generative deep learning models.

More topics of Coffee Journal Club...

Diese Website nutzt Cookies, um Ihr Nutzererlebnis zu verbessern. Durch die weitere Nutzung dieser Webseite erklären Sie sich mit der Verwendung von Cookies einverstanden. Weitere Informationen hierzu finden Sie in unserer Datenschutzerklärung