Voice Over WMSN

Funded Research Project (Turkcell)
Year: 2011-2012 Project Team: Şebnem Baydere (sbaydere [at] cse.yeditepe.edu.tr), Okan Türkeş.  

Multimedia data provide versatility to any kind of application concerned with information extraction, by rendering a phenomenon down to processable data. With the utilization of ubiquitous content forms in Wireless Sensor Networks (WSNs), the era of Wireless Multimedia Sensor Networks (WMSNs) is triggered. However, under-resourced sensor nodes of these networks have to struggle with multiple data features. Audio encoding techniques matter to sustain a certain content validity to be preserved in the lossy nature of a WSN. Besides, handling bulk data gains importance in terms of computation and traffic overhead. Basic characteristics of a voice signal must be compromised with network properties to maintain an admissible quality. This  project analyzes voice coding and transmission in WMSNs in terms of data and network qualifications. With a comprehensive examination of the existing schemes relating to WMSNs and consideration of mass data delivery requirements in the shade of system constraints, several contributions are made: A priority-based data coding scheme is proposed, with a set of error-resilience methods. For this scheme, a voice transmission framework (VTF) is presented which deals with the robust delivery of partitions. A wide range of experiments over a data set are conducted in a real testbed environment with several network properties. A simulation is devised to weigh an affordable trade-off between data quality and network capabilities. Experiments are assessed with a modified version of an objective transmission rating factor which can be mapped to a certain perceptibility. With these attributes, VTF offers a promising quality improvement for WMSN applications.

Permanent link to this article: http://cse.yeditepe.edu.tr/en/research/projects/funded-research-projects/voice-over-wmsn/