Аннотация
The recognition of Activities of Daily Living (ADLs) is a fundamental component of Ambient Assisted Living (AAL) systems, facilitating the continuous monitoring of elderly individuals to promote their well-being. The accurate identification of routine activities, such as walking, sitting, and sleeping, is essential for detecting deviations indicative of potential health risks, including falls. This study proposes an advanced method to enhance the accuracy and reliability of ADL recognition by incorporating temporal logic and model checking. Temporal logic effectively represents the sequential dependencies of activities, while model checking ensures adherence to predefined temporal constraints, thereby improving system robustness. The proposed framework integrates multimodal sensor data from wearable and environmental sources to enable real-time ADL detection. Experimental evaluations conducted on publicly available datasets demonstrate a recognition accuracy of 91%, outperforming conventional approaches by 12%. Furthermore, model checking achieves a 94% success rate in validating temporal compliance. The findings highlight the efficacy of the proposed approach in providing a structured and reliable solution for real-time ADL detection in AAL environments. Future research directions include the integration of deep learning methodologies to address more complex activity patterns.