Physical Activity and Behaviour Monitoring

Overview

Physical Activity and Behaviour Monitoring is one of the cornerstones of the work undertaken at the Brain and Behaviour Lab. Both physical activity and behaviour are intrinsically linked to health and it is our belief that if these can be monitored in an unobtrusive, continuous manner a range of health insight can be generated. In support of this we have a number of research initiatives underway to investigate these aspects of health care.

Physical Activity relates to the physical movement of the human body in the world; the way in which the user moves and the postures they are in. As such it is concerned with fine time-granularity, of the order of seconds to minutes, and specifically with physical motion. In contrast, Behaviour relates to longitudinal actions and includes not just the action, but the surrounding context. Indeed, behaviour incorporates physical activity but includes much more information and is concerned with the longer duration, stretching often into months or even longer.

Physical Activity Monitoring

Physical Activity is most commonly monitored with accelerometers, gyroscopes and magnetometers, which, when used in conjunction, are referred to as inertial measurement units (IMUs). It is not necessary to include all three sensors, and often the magnetometer and/or the gyroscope can be omitted, with the most crucial data being provided by the accelerometer, which measures the acceleration (and therefore movement) of the object to which it is attached—for our purposes this is most often a human user, but can be an object such as a wheelchair or prosthetic limb.

Data from an IMU provides data on the physical motion of the thing it is attached to. This in turn provides information about what is happening to that thing. In the case of a human user with the IMU mounted on the arm it may be possible to detect a periodic swinging motion and infer that walking is taking place. It is also possible from the use of an IMU to estimate total physical activity of the user, and from this to estimate their calorific energy expenditure related to their physical activity. Data on energy expenditure has uses in a number of fields and the creation of an accurate estimate of this from wearable IMUs is an active research topic in the Brain and Behaviour Lab.

Behaviour Monitoring and Multi-Sensor Data Fusion

While physical activity monitoring is concerned with physical motion and posture over relatively short timescales, behaviour monitoring is concerned with much longer timescales, typically in the order of weeks, months or even ongoing in perpetuity. A person’s behaviour is comprised primarily from what they are doing; when they are doing it; the place and environment in which they are. In addition to these three components of behaviour, the act of repetition is a key factor in behaviour analysis as repeated actions in the same place at the same time help to establish patterns of behaviour for an individual.

Physical activity monitoring can go a long way towards establishing the moment by moment “what” and “when” of behaviour, the “where” (meaning physical location and environmental contexts) requires other sensors to determine. Location can be detected using GPS, for outdoors; or passive infra-red or radio frequency techniques for indoors. Other environmental contextual information can be acquired using a sensor network in the environment in question. Sensors in this network might be used to detect, for example, electricity usage, kitchen appliance use, or bed occupancy. Contextual information can also be acquired through user self reports and this is an effective method to generate data that can be hard to directly observe.

The data gathered by the behavioural monitoring system are typically stored in a central database and used in a multi-sensor data fusion system to extract patterns of behaviour that are relevant to the user. This layer is often customised to meet a particular set of user requirements and/or to be tailored to a specific disease or condition. Extracted behavioural patterns typically focus on identifying normal patterns of behaviour for a user, identifying deviations from these patterns and predicting the impact on the user’s health of changes in their behaviour. This area of research is one of the ongoing themes of the Brain and Behaviour Lab.