本論文結合經驗模態分解於壓力訊號處理與居家活動識別。經驗模態分解技術可以將活動壓力訊號分解成數個內建模態函數,並將一些特定的模態函數合成,形成一個有特定活動特徵的壓力訊號。相對於原始未處理過之信號,基於本論文上述處理後之訊號,於相同之壓力訊號處理過程中,特定活動的識別率有明顯提升。本論文提出了壓力地板的活動識別系統,此系統能讓使用者在家中的任何地方,無需額外配戴或是懸掛任何活動感測器,就能達到家人與居家服務系統溝通的目的。因此一新技術,居家服務系統能了解家人的活動現況,據此即可進一步提供智慧與舒適的服務。本居家活動識別系統分為五個主要的部分:經驗模態分解、居家活動識別、活動特徵萃取、活動建模與活動識別。此系統可以使一些特定活動的識別率平均提升約67%,所以此系統在居家照護上,可以提供更聰明、準確與適切的服務。
This thesis integrates the empirical mode decomposition into the pressure signal processing and living activities identification. The empirical mode decomposition can decompose the activity pressure signal into a number of intrinsic mode functions, and combine some specified intrinsic mode functions to form an abstracted pressure signal with obvious characteristics of the specified activity. Based on above abstracted pressure signal, the identification rates of the specified activities are enhanced relative to the original signal after the same processing in this thesis. This thesis also presents a living activities identification system with pressure-sensed floors, this system make families to communicate with the home-cared serving systems at any place in home without wearing or hanging any motion sensors. In addition, this identification system also makes the serving system to understand the activity situation of the families, and to offer intelligence and conformable services further. This living activities identification system is composed by five major parts: empirical mode decomposition, activity signal processing, activity feature extraction, activity modeling, and activity identification. This integration can promotes some specified activities’ identification rate about 67% in average, so it can make the home-cared systems to offer services more smartly, accurately and appropriately.