Human Behavior Understanding

In this paper the acceptance of a sensor-based Ambient Assisted Living (AAL) application is investigated. To get an insight into the users’ perception and needs, three fictive use scenarios were created that illustrated the potential features of the technology. Consequently, the scenarios were presented to primary (i.e., older adults) and secondary (i.e., formal and informal caregivers) user groups. Through focus groups and semi-structured interviews in France, UK and Belgium, fourteen design implications could be identified based on the preliminary analyses of the users’ feedback. These implications will direct the future testing and development of the conceptual technology and can be meaningful to related AAL applications.

Nearly half of the senior citizens report difficulty initiating and maintaining sleep. Frequent visits to the bathroom in the middle of the night is considered as one of the major reasons for sleep disorder. This leads to serious diseases such as depression and diabetes. In this paper, we propose to use a network of cheap low-resolution visual sensors (30 \(\times \) 30 pixels) for long-term activity analysis of a senior citizen in a service flat. The main focus of our research is on elderly behaviour analysis to detect health deterioration. Specifically, this paper treats the analysis of sleep patterns. Firstly, motion patterns are detected. Then, a rule-based approach on the motion patterns is proposed to determine the wake up time and sleep time. The nightly bathroom visit is identified using a classification-based model. In our evaluation, we performed experiments on 10 months of real-life data. The ground truth is collected from the diaries in which the senior citizen wrote down his sleep time and wake up time. The results show accurate extraction of the sleep durations with an overall Mean Absolute Error (MAE) of 22.91 min and Spearman correlation coefficient of 0.69. Finally, the nightly bathroom visits analysis indicate sleep disorder in several nights.

Technology designed to sense behavior, often neglects to directly incorporate subjective input from (elderly) users. This paper presents experiences in deploying technology that considers the elderly user and their subjective input as a way to enrich sensor data systems and empower the user. For this purpose, the paper draws on: (1) Observations of shortcomings in terms of capturing objective data from sensors as experienced in long-term deployments in the homes of older adults; (2) The design and evaluation of a wide range of applications especially designed to enable older adults to give subjective input on how they are doing, including an interactive television quiz, a talking picture frame and a tangible mood board, and (3) The development and field study of one application, the ‘Mood button’ in particular, that was tested in real-world sensing settings to work with a commercial sensing system. In doing this, this work aims to contribute towards successful sensing deployments and tools that give more control to the (elderly) end-user.

Functional assessment is the test of the ability of a person to perform basic self-care activities that are instrumental for living safely and independently in a home. Gerontology classifies these self-care activities as Activities of Daily Living (ADL). There exist many clinical and systems measures for performing functional assessment. This paper critically reviews the state of art in these assessments. This paper also talks about the disconnect between the clinical and the technological measures. It also discusses future directions to establish a practical and objective method of conducting functional assessments.

Smartphone notifications are often delivered without considering user interruptibility, potentially causing frustration for the recipient. Therefore research in this area has concerned finding contexts where interruptions are better received. The typical convention for monitoring interruption behaviour assumes binary actions, where a response is either completed or not at all. However, in reality a user may partially respond to an interruption, such as reacting to an audible alert or exploring which application caused it. Consequently we present a multi-step model of interruptibility that allows assessment of both partial and complete notification responses. Through a 6-month in-the-wild case study of 11,346 to-do list reminders from 93 users, we find support for reducing false-negative classification of interruptibility. Additionally, we find that different response behaviour is correlated with different contexts and that these behaviours are predictable with similar accuracy to complete responses.

Fatigue is one of the symptoms associated with frailty among older adults, a syndrome that signals a progressive deterioration of physical and mental capacity. It is commonly measured by clinical studies, physical tests, or questionnaires, involving effort some people can’t attain. We describe a method to assess the fatigue a person experiences while walking by measuring their demand for oxygen and heart rate. The approach makes use of the accelerometer in the mobile phone, thus allowing for continuous fatigue assessment under naturalistic conditions. Results show a margin of error of less than 5 heart beats per minute and no significant difference between heart rate’s base-line fatigue classification and predicted fatigue classification; which compares favourably with current approaches. The method can be used to monitor health conditions among the frail.

Current advances in location tracking technology provide exceptional amount of data about the users’ movements. The volume of geospatial data collected from moving users’ challenges human ability to analyze the stream of input data. Therefore, new methods for online mining of moving object data are required. One of the popular approaches available for moving objects is the prediction of the unknown future location of an object. In this paper we present a new method for online prediction of users’ next important locations to be visited that not only learns incrementally the users’ habits, but also detects and supports the drifts in their patterns. Our original contribution includes a new algorithm of online mining association rules that support the concept drift.

The automatic detection of behavioral changes in older adults living in geriatric centers is relevant for physicians and caregivers. These changes could indicate an incipient symptom of a disease or a steep decline in the health of the person. Abnormal pattern discovery has been studied in the context of an array of (wearable) sensors (i.e. accelerometers, infrared, cameras, etc.) dedicated to monitor the older adult. In this work we explore the use of manually annotated records, the type of records maintained by caregivers in a daily log. These annotations have low semantic value, and consist in a sequence of keywords about the activity being carried out by the older adult. This information is often overseen because it could be noisy, incomplete and redundant. We tested a data-driven approach to identify patterns from daily activity records, which were collected over six months from a group of older adults in a geriatric center. The results show that through simple data processing techniques it is posible to identify abnormal patterns in daily activities associated with behavioral changes over time.

Human behavior learning plays an important role in ambient assisted living since it enables service personalization. Current work in human behavior learning do not consider the context under which a behavior occurs, which hides some behaviors that are frequent only under certain conditions. In this work, we present the notion of a contextualized behavior pattern, which describes a behavior pattern with the context in which it occurs (i.e. nap when raining) and propose an algorithm for finding these patterns in a data stream. This is our main contribution. These patterns help to better understand the routine of a user in a smart environment, as is evidenced when testing with a public dataset. This algorithm could be used to learn behaviors from users in an ambient assisted living environment in order to send alarms when behavior changes occur.

We describe stroke patients’ activity patterns and trends based on motion data acquired during their stay in an ambulatory day-care centre. Our aim was to explore and quantify intensity and development in the patients’ activity patterns as these may change during the rehabilitation process. We analyse motion data recordings from wearable inertial measurement units of eleven patients up to eleven days, totally 102 recording days. Using logic rules, we extract activity primitives, including affected arm move, sit, stand, walking, etc. from selected channels of the continuous median-filtered sensor data. Using relative duration of the activity primitives, we examine patient activity patterns regarding independence in mobility, distribution of walking over the days and trends in using the affected body side. Due to the heterogeneity of patients’ behaviour, we focused on analysing patient-specific activity patterns. Our exploration showed that the rule-based activity primitive analysis is beneficial to understand individual patient activity.

Caregivers of people with cognitive impairment need assistive technologies capable of reducing stress of constant monitoring of patient. In this paper we discuss information technologies employed for sensing and identification of basic behavioral activities of a patient in a low-cost caregiver assisting system. By analyzing readings from heterogeneous sensors, the system automatically detects the activities, assesses risks which they may have for the patient’s health, evaluates emergency of assistance and alerts the caregiver in a case of emergency. We present algorithms for activity identification, emergency computation and show results of empirical evaluation in a prototype in-home caregiver assisting system. As experiments revealed, the system has identification rate for basic activities higher than 94 %.