Pierluigi Casale

Human activity recognition from accelerometer data using a wearable device

Activity Recognition is an emerging field of research, born from the larger fields of ubiquitous computing, context-aware computing and multimedia. Recently, recognizing everyday life activities becomes one of the challenges for pervasive computing. In our work, we developed a novel wearable system easy to use and comfortable to bring. Our wearable system is based on a new set of 20 computationally efficient features and the Random Forest classifier. We obtain very encouraging results with classification accuracy of human activities recognition of up to 94%.

Personalization and user verification in wearable systems using biometric walking patterns

In this article, a novel technique for user’s authentication and verification using gait as a biometric unobtrusive pattern is proposed. The method is based on a two stages pipeline. First, a general activity recognition classifier is personalized for an specific user using a small sample of her/his walking pattern. As a result, the system is much more selective with respect to the new walking pattern. A second stage verifies whether the user is an authorized one or not. This stage is defined as a one-class classification problem. In order to solve this problem, a four-layer architecture is built around the geometric concept of convex hull. This architecture allows to improve robustness to outliers, modeling non-convex shapes, and to take into account temporal coherence information. Two different scenarios are proposed as validation with two different wearable systems. First, a custom high-performance wearable system is built and used in a free environment. A second dataset is acquired from an Android-based commercial device in a ‘wild’ scenario with rough terrains, adversarial conditions, crowded places and obstacles. Results on both systems and datasets are very promising, reducing the verification error rates by an order of magnitude with respect to the state-of-the-art technologies.

Approximate polytope ensemble for one-class classification

In this work, a new one-class classification ensemble strategy called approximate polytope ensemble is presented. The main contribution of the paper is threefold. First, the geometrical concept of convex hull is used to define the boundary of the target class defining the problem. Expansions and contractions of this geometrical structure are introduced in order to avoid over-fitting. Second, the decision whether a point belongs to the convex hull model in high dimensional spaces is approximated by means of random projections and an ensemble decision process. Finally, a tiling strategy is proposed in order to model non-convex structures. Experimental results show that the proposed strategy is significantly better than state of the art one-class classification methods on over 200 datasets. HighlightsThe methodology uses a convex-hull for modeling one-class classification problems.Random projections are used to approximate the convex-hull in high dimensional spaces.Expansions of the approximate hulls are considered to set the optimal operating point.Exhaustive validation is performed on three different typologies of problems.

Approximate convex hulls family for one-class classification

In this work, a new method for one-class classification based on the Convex Hull geometric structure is proposed. The new method creates a family of convex hulls able to fit the geometrical shape of the training points. The increased computational cost due to the creation of the convex hull in multiple dimensions is circumvented using random projections. This provides an approximation of the original structure with multiple bi-dimensional views. In the projection planes, a mechanism for noisy points rejection has also been elaborated and evaluated. Results show that the approach performs considerably well with respect to the state the art in one-class classification.

Cardiorespiratory fitness estimation in free-living using wearable sensors

OBJECTIVE In this paper we propose artificial intelligence methods to estimate cardiorespiratory fitness (CRF) in free-living using wearable sensor data. METHODS Our methods rely on a computational framework able to contextualize heart rate (HR) in free-living, and use context-specific HR as predictor of CRF without need for laboratory tests. In particular, we propose three estimation steps. Initially, we recognize activity primitives using accelerometer and location data. Using topic models, we group activity primitives and derive activities composites. We subsequently rank activity composites, and analyze the relation between ranked activity composites and CRF across individuals. Finally, HR data in specific activity primitives and composites is used as predictor in a hierarchical Bayesian regression model to estimate CRF level from the participants habitual behavior in free-living. RESULTS We show that by combining activity primitives and activity composites the proposed framework can adapt to the user and context, and outperforms other CRF estimation models, reducing estimation error between 10.3% and 22.6% on a study population of 46 participants. CONCLUSIONS Our investigation showed that HR can be contextualized in free-living using activity primitives and activity composites and robust CRF estimation in free-living is feasible.

Identifying Physical Activity Profiles in COPD Patients Using Topic Models

With the growing amount of physical activity (PA) measures, the need for methods and algorithms that automatically analyze and interpret unannotated data increases. In this paper, PA is seen as a combination of multimodal constructs that can cooccur in different ways and proportions during the day. The design of a methodology able to integrate and analyze them is discussed, and its operation is illustrated by applying it to a dataset comprising data from COPD patients and healthy subjects acquired in daily life. The method encompasses different stages. The first stage is a completely automated method of labeling low-level multimodal PA measures. The information contained in the PA labels are further structured using topic modeling techniques, a machine learning method from the text processing community. The topic modeling discovers the main themes that pervade a large set of data. In our case, topic models discover PA routines that are active in the assessed days of the subjects under study. Applying the designed algorithm to our data provides new learnings and insights. As expected, the algorithm discovers that PA routines for COPD patients and healthy subjects are substantially different regarding their composition and moments in time in which transitions occur. Furthermore, it shows consistent trends relating to disease severity as measured by standard clinical practice.

Cardiorespiratory fitness estimation using wearable sensors: laboratory and free-living analysis of context-specific submaximal heart rates.

In this work, we propose to use pattern recognition methods to determine submaximal heart rate (HR) during specific contexts, such as walking at a certain speed, using wearable sensors in free living, and using context-specific HR to estimate cardiorespiratory fitness (CRF). CRF of 51 participants was assessed by a maximal exertion test (V̇o2 max). Participants wore a combined accelerometer and HR monitor during a laboratory-based simulation of activities of daily living and for 2 wk in free living. Anthropometrics, HR while lying down, and walking at predefined speeds in laboratory settings were used to estimate CRF. Explained variance (R(2)) was 0.64 for anthropometrics, and increased up to 0.74 for context-specific HR (0.73-0.78 when including fat-free mass). Next, we developed activity recognition and walking speed estimation algorithms to determine the same contexts (i.e., lying down and walking) in free living. Context-specific HR in free living was highly correlated with laboratory measurements (Pearsons r = 0.71-0.75). R(2) for CRF estimation was 0.65 when anthropometrics were used as predictors, and increased up to 0.77 when including free-living context-specific HR (i.e., HR while walking at 5.5 km/h). R(2) varied between 0.73 and 0.80 when including fat-free mass among the predictors. Root mean-square error was reduced from 354.7 to 281.0 ml/min by the inclusion of context-specific HR parameters (21% error reduction). We conclude that pattern recognition techniques can be used to contextualize HR in free living and estimated CRF with accuracy comparable to what can be obtained with laboratory measurements of HR response to walking.

Personalization of Energy Expenditure Estimation in Free Living Using Topic Models

We introduce an approach to personalize energy expenditure (EE) estimates in free living. First, we use topic models to discover activity composites from recognized activity primitives and stay regions in daily living data. Subsequently, we determine activity composites that are relevant to contextualize heart rate (HR). Activity composites were ranked and analyzed to optimize the correlation to HR normalization parameters. Finally, individual-specific HR normalization parameters were used to normalize HR. Normalized HR was then included in activity-specific regression models to estimate EE. Our HR normalization minimizes the effect of individual fitness differences from entering in EE regression models. By estimating HR normalization parameters in free living, our approach avoids dedicated individual calibration or laboratory tests. In a combined free-living and laboratory study dataset, including 34 healthy volunteers, we show that HR normalization in 14-day free-living data improves accuracy compared to no normalization and normalization based on activity primitives only (29.4% and 19.8 % error reduction against lab reference). Based on acceleration and HR, both recorded from a necklace, and GPS acquired from a smartphone, EE estimation error was reduced by 10.7% in a leave-one-participant-out analysis.

Face-to-Face Social Activity Detection Using Data Collected with a Wearable Device

In this work the feasibility of building a socially aware badge that learns from user activities is explored. A wearable multisensor device has been prototyped for collecting data about user movements and photos of the environment where the user acts. Using motion data, speaking and other activities have been classified. Images have been analysed in order to complement motion data and help for the detection of social behaviours. A face detector and an activity classifier are both used for detecting if users have a social activity in the time they worn the device. Good results encourage the improvement of the system at both hardware and software level.

Comparison of Machine Learning Techniques for Psychophysiological Stress Detection

Previous research has indicated that physiological signals can be used to detect mental stress. There is however no consensus on the optimal algorithm for this detection. The aim of this study is to compare different machine learning techniques for the measurement of stress based on physiological responses in a controlled environment. Electrocardiogram (ECG), galvanic skin response (GSR), temperature and respiration were measured during a laboratory stress test. Six machine learning techniques were investigated using a general and personal approach. The results show that personalized dynamic Bayesian networks and generalized support vector machines render the best average classification results with 84.6 % and 82.7 % respectively.