• Wrist-Worn sensor
• Wearable Sensor
• Pervasive Health-Care
• Gait Detection

In recent years, the widespread adoption of wearable sensors has led to the study of innovative technological solutions to improve user's health and well-being.
In this context, a particularly relevant application is the analysis of human walking patterns (gait analysis). Walking is one of the most common human activities, and thus it can be frequently detected during daily routines. Gait patterns, in turn, provide useful information for a number of applications.
For example, wearable-based gait analysis has been proposed as an unobtrusive alternative to clinical assessment in order to early detect health-related conditions that affect motor ability.
Indeed, wearable-based gait analysis can be used to continuously monitor the user gait patterns and promptly detect anomalies that may indicate the presence of a health-related condition. Such early detection method, in turn, may be exploited to take appropriate preemptive measures and minimize the risk of adverse health outcomes.
Wrist-worn devices, such as smartwatches and smart bands, have brought about the unprecedented opportunity to continuously monitor gait during daily routines. However, the use of a single wrist-worn unit for gait analysis is challenging for a variety of reasons and has been scarcely investigated in the literature. One relevant issue with wrist-worn devices is the reliable detection and extraction of gait segments (i.e., sensor samples containing gait data), which is the essential building block of a gait analysis system. Indeed, the signal collected at the user's wrist is subject to a significant "noise" with respect to other body positions (e.g. waist), mainly due to the arm swing while walking and other unpredictable hand movements.
The aim of this thesis is to investigate the design and evaluation of a lightweight and reliable gait detection technique for wrist-worn devices. To this end, the proposed method creates a personalised model of the user's gait patterns. The model is created through an unsupervised training phase, which requires the temporary use of an additional device (smartphone) to properly label the acquired acceleration samples as gait or hand movements. After, anomaly detection is used to distinguish gait from other activities. Gait data from 20 volunteers has been collected to test and evaluate the proposed technique. Volunteers were asked to walk at different pace, with their normal hand swing or placing the hand inside of a pocket. Results show that the proposed method can reliably distinguish gait from hand movements. This will foster the development of trustworthy gait analysis applications based on a wrist-worn device.