Context-Aware Data Analytics for Sensor Arrays

点击翻译按钮获取中文摘要

The main objective of this research program is to design smart context-aware data processing algorithms to extract useful information from large data streams. We will use context information from the environment to improve the system’s performance. During the past few years, advances in the areas of semiconductors and micro electro mechanical systems (MEMS) led to the introduction of new compact sensors such as pressure sensors, accelerometers, gyroscopes, microphone arrays and electronic noses. These advances, coupled with powerful processors and embedded wireless interfaces enabled the introduction of smart sensor systems with minimal power consumption. This project will propose novel noise cancellation, feature extraction, pattern recognition and classification approaches to extract useful information from multiple sensors. These sensors must work synergistically to provide the user with the required information at the right time, and in the right format given a specific context. The project will consider fixed sensors and smart-phone-based sensors for two applications: the design of smart homes and audio signal processing. Fixed sensors are usually deployed in a given environment such as a smart home or critical infrastructure. In this program, we will investigate several sensor technologies, in particular pressure-sensitive arrays and microphone arrays. Pressure-sensitive arrays measure the pressure applied at any point on their surface. When embedded into floors, these sensors are capable of detecting intruders, or performing gait analysis for biometric purposes or health diagnosis. When placed under a bed mattress, advanced signal processing algorithms can be used to extract bed occupancy, breathing, or bed exit characteristics (timing, symmetry, bouncing or centre of pressure trajectory). During the past few years, we proposed several algorithms to extract the timing and symmetry information. More research is needed to extract the bouncing and centre of pressure trajectory. We will also propose approaches to auto calibrate these systems for any individuals regardless of their weight and type of mattress used. Microphone arrays are able to localize and track various speakers, thus removing the undesired background noise and interference. Audio classification algorithms will be used to classify various speakers and perform important security and health monitoring tasks. Smart phones and tablets can play an important role in data acquisition, data transmission, context awareness and information presentation to the users. This project will use the internal sensors of a smart phone to monitor daily living activity. It will also use the phone to collect data from various sensors using their bluetooth or wifi communication links, and send it to the appropriate servers through the wifi or cellular networks. Smart phones can provide important context awareness information. For example, the internal microphone can characterize the ambient acoustic environment and classify various talkers. The internal ambient light sensor and temperature sensor can provide additional context information while the internal clock can provide the required synchronization. The accelerometers, gyroscope and GPS could monitor physical activity and exercise compliance. The next generation of smart phones will likely be equipped with more advanced sensors such as smell sensors. Smart phones or tablets can also perform some pre-processing and could convey security risks to authorities or clinical information to healthcare professionals. In summary, this project will propose novel algorithms for processing multiple data streams to extract specific information and present it to the users at the appropriate level in the appropriate format.