SBIR-STTR Award

We propose an analog transmission of a speech cepstrum for the dual purposes of bandwidth reduction and improvements in speech quality under extremely challenging channel environments. By utilizing the speech cepstrum, we exploit a well-studied and widely-used sparse representation of speech. The novelty in the method is that we propose to “spread” the waveform defined by the first few cepstral coefficients over the speech frame length, i.e. time-domain spreading. This spread cepstrum approach reduces the bandwidth and may offer robustness to noise for low-power applications.

Keywords:
Analog Transmission Of Speech, Low Bandwidth, Spread Cepstrum Representation, Improved Intelligibility Of Speech, Vocoders, First Responder, Tactical Communications

Project Pika addresses the challenge of receiving messages on a radio channel when multiple users inadvertently collide on the channel—an event that is especially likely during times of disaster or emergency. In Phase I we demonstrated the feasibility of this signal source separation through analysis and simulation for a subset of communication modes under the assumption of certain constraints. In Phase II, we relax these constraints and generalize the capability for more communication waveforms and signal sources, and develop a prototype communication system demonstrating the capability in open air tests through real world propagation. Of particular emphasis is the underdetermined case, where there are more signals in a received mixture than available receiver channels. The project will result in receiver software that is portable to modern software defined radio systems, and that is compatible with current hardware and legacy waveforms.

Keywords:
Blind Source Separation (Bss), Modulation Recognition, Radio Communications, First Responder, Interference Mitigation, Software Defined Radio (Sdr), Spectrum Management, Joint