Wireless Multimedia Sensor Networks (WMSNs) are widely deployed to meet the ever-growing demands for surveillance. Compared with Wireless Sensor Networks (WSNs), WMSNs require higher throughput and real-time transmission, which introduces new challenges to sensor node architecture design. This thesis focuses on voice sensor nodes design using ZigBee technology.After the analysis of WMSNs architecture and review of representative voice sensor node schemes, a ZigBee System-on-Chip (SoC) based voice acquisition and transmission solution is presented in this thesis, aiming at reducing the cost and power consumption. An external mono CODEC is also adopted to improve the voice acqusition quality. In the aspect of software design, we compile the hardware drivers of the data collection, processing, and transmission modules, as well as the single-hop wireless communication programs. With the consideration of the inherent multidimensionalism characteristics of WMSNs, an extended IEEE 802.15.4 MAC frame format is proposed to classify the multimedia data. Furthermore, self-adaptive frequency agility is applied to overcome the QoS drop problem caused by channel instability. Our experiment results demonstrate that the proposed voice sensor nodes can achieve a high packet delivery rate of 90% within a maximum distance of 25 meters, even in a harsh communication environment. |