A Method for Robust Online Classification using Dictionary Learning: Development and Assessment for Monitoring Manual Material Handling Activities Using Wearable Sensors

With the rapid development of sensor technology in recent years, there has been a growing need to quickly and accurately analyze sensor data, and to make decisions online (potentially even in real-time). This need exists in many different industry sectors. One illustrative application domain is healthcare: for example, wearable sensors can be integrated with online decision-making algorithms to help elderly patients who need continuous care [23]. Another domain is in manufacturing, where there is an ongoing and critical need to monitor part quality using sensor data [24, 25, 26]. For workers in several domains who are engaged in manual material handling (MMH), the risks of musculoskeletal injury are relatively high and such risks are associated with specific work methods and exposure duration [27, 28]. For such a case, applications of wearable sensors for MMH online monitoring have the potential to be an effective means to monitor the status of the workers' operational conditions (e.g., physical demands imposed, performed task characteristics), based on which online decision making can be appropriately performed [29]. A. The sparse signal reconstruction problem In this section, we briefly review sparse signal reconstruction methods, including the general problem, and the least absolute shrinkage and selection operator (LASSO) [30] method, which are both directly related to the new approach proposed in this paper.