Looking at the subject more broadly, the recent doctor says it is important to recognise that the lives of modern people are very different from the lives of people in the past. Physical exertion is no longer a prerequisite for everyday life and modern comfortable living conditions allow people to cope with an unprecedented lack of physical activity. However, this does not mean that this is a good lifestyle.

Active lifestyle has a positive impact on health

In fact, the difference between our modern lifestyle and our genetic makeup has important health implications, such as for bone density, cardiovascular diseases, obesity, body composition, and insulin resistance. It is therefore highly important to promote an active lifestyle according to Allik – studies confirm that routine physical activity has a positive effect, reducing the risk of diabetes, cardiovascular diseases, and obesity and increasing mental well-being.

Physical activity and training can be tracked using wearable devices that measure, collect, and analyse the physiological data of the user. As this kind of monitoring makes exercise more interactive and allows users to get a better overview of their progress, it often motivates users to adopt a more active lifestyle.

The equipment has to be small and unobtrusive

‘The aim of my doctoral thesis was to promote novel methods for assessing physical fatigue and monitoring human activity that could be implemented in real time in various wearable systems,’ explains Ardo Allik. ‘With the advancement of technology, devices and applications that help users track their physical activity, exercise schedules, exercises, and calories burnt have gained popularity. For convenient use, these solutions must be small and unnoticeable, which in turn creates a significant requirement for the optimisation of various aspects of these systems.’

Three main objectives of the thesis

Allik stresses that his doctoral thesis had three major objectives. One of the main objectives was to improve the detection of physical activity, which provides real-time information about the movements of the user. The research optimised the length of the classification window, the sampling rate of the accelerometer, and the selection of characteristics. Real-time motion detection provides valuable information to improve the quality of feedback from activity monitors or to provide additional safety by monitoring the status of users working in high-risk environments.

Another important objective was to evaluate different formulas for estimating baseline material costs to promote energy expenditure estimation. In particular, energy expenditure is an important parameter for physical activity studies and is often used as a correlate. It is also very useful for adjusting the dietary habits of people or assessing the health of a larger population.

The third major goal was to create a method for the real-time assessment of physical fatigue suitable for wearable devices using real-time and easily measurable cardiovascular parameters. In this context, fatigue refers to an altered physiological state resulting in a decline in mental and physical abilities. This change is also reflected in the values of various physiological parameters’.

Why is a high-quality assessment of fatigue important? According to Allik, one of the reasons why it is necessary to be able to correctly and qualitatively assess fatigue is that people experience fatigue not only when exercising, but also at work or when performing specific tasks, which in turn increases the risk of accidents at work, directly affecting people’s mental and physical ability to perform even simple activities.

Results of the doctoral research

As a result of the work, the length of the classification window as well as the discretisation or sampling frequency of the accelerometer were optimised; the number of characteristics was reduced from 110 to 13 using different methods for choosing characteristics without lowering the classification result.

‘The set of parameters analysed in the doctoral thesis is able to selectively distinguish between resting and physically fatigued states. The machine learning-based model demonstrated the ability to distinguish between different fatigue states, i.e. in less-fatigued people who were less affected by physical exercise and in highly fatigued people,’ explains Allik. It was found that the Mifflin–St Jeor model performed best in estimating energy expenditure, with a smaller root mean square deviation of 175 kcal compared to the measured value.

In addition, in his work, Allik proposed a number of ways to further develop modern physical fatigue assessment and motion detection methods by improving their performance or optimising them for use in wearable systems.

Read the PhD thesis by Ardo Allik, ‘Advancing Novel Physical Fatigue Assessment and Human Activity Monitoring Methods towards Perzonalized Feedback with Wearable Sensors’ in the digital collection of Tallinn University of Technology.