End to End Deep Neural Networks Radio Receiver for Speech Signals

 Frequency modulation (FM) is a form of radio broadcasting which is widely used nowadays and has been in use for over a century, since 1906 when the first radio audio broadcast was made.  FM performance is severely degraded when the system noise amount exceeds a critical level, above which the traditional FM receiver breaks.  Traditional methods used in radio receivers divide the problem into two parts – signal demodulation and speech enhancement.  We suggest a software-defined-radio receiver that detects and enhances speech, by adopting an end-to-end learning based approach and utilizing the prior information of transmitted speech message in the demodulation process.  The new system yields high performance detection for both acoustical disturbances and communication channel noise