Ramyar Saeedi

Power-Aware Computing in Wearable Sensor Networks: An Optimal Feature Selection

Wearable sensory devices are becoming the enabling technology for many applications in healthcare and well-being, where computational elements are tightly coupled with the human body to monitor specific events about their subjects. Classification algorithms are the most commonly used machine learning modules that detect events of interest in these systems. The use of accurate and resource-efficient classification algorithms is of key importance because wearable nodes operate on limited resources on one hand and intend to recognize critical events (e.g., falls) on the other hand. These algorithms are used to map statistical features extracted from physiological signals onto different states such as health status of a patient or type of activity performed by a subject. Conventionally selected features may lead to rapid battery depletion, mainly due to the absence of computing complexity criterion while selecting prominent features. In this paper, we introduce the notion of power-aware feature selection, which aims at minimizing energy consumption of the data processing for classification applications such as action recognition. Our approach takes into consideration the energy cost of individual features that are calculated in real-time. A graph model is introduced to represent correlation and computing complexity of the features. The problem is formulated using integer programming and a greedy approximation is presented to select the features in a power-efficient manner. Experimental results on thirty channels of activity data collected from real subjects demonstrate that our approach can significantly reduce energy consumption of the computing module, resulting in more than 30 percent energy savings while achieving 96.7 percent classification accuracy.

Cost-sensitive feature selection for on-body sensor localization

Activity recognition systems have demonstrated potential in a broad range of applications. A crucial aspect of creating large scale human activity sensing corpus is to develop algorithms that perform activity recognition in a way that users are not limited to wear sensors on predefined locations on the body. Therefore, effective on-body sensor localization algorithms are needed to detect the location of wearable sensors automatically and in real-time. However, power optimization is a major concern in the design of these systems. Frequent need to charge multiple sensor nodes imposes much burden on the end-users. In this paper, we propose a novel signal processing approach that leverages feature selection algorithms to minimize power consumption of node localization. With the real data collected using wearable motion sensors, we demonstrate that the proposed approach achieves an energy saving that ranges from 88% to 99.59% while obtaining an accuracy performance between 73.15% and 99.85%.

Patient-centric on-body sensor localization in smart health systems

A major obstacle in widespread adoption of current wearable monitoring systems is that sensors must be worn on predefined locations on the body. In order to continuously detect sensor locations, we propose a localization algorithm that allows patients to wear the sensors on different body locations without having to adhere to a specific installation protocol. Our approach achieves localization accuracy of 90.8% even when the sensor nodes are mis-oriented. Integration of the resulting location information as a feature in an activity recognition classifier significantly increased the recognition accuracy from 23.5% to 99.5%.

Impact of sensor misplacement on estimating metabolic equivalent of task with wearables

Metabolic equivalent of task (MET) indicates the intensity of physical activities. This measurement is used in providing physical activity intervention in many chronic illnesses such as coronary heart disease, type-2 diabetes, and cancer. Due to the small size, portability, low power consumption, and low cost, wearable motion sensors are widely used to estimate MET values. However, one major obstacle in widespread adoption of current wearable monitoring systems is that the sensors must be worn on predefined locations on the body. This imposes much discomfort for users as they are not allowed to wear the sensors on their own desired body locations. In addition, non-adherence to the predefined location of the sensors results in significant reduction in the accuracy of physical activity monitoring. In this paper, we propose a framework for sensor location-independent MET estimation. We introduce a sensor localization approach that allows users to wear the sensors on different body locations without having to adhere to a specific installation protocol. We study how such an algorithm impacts the performance of MET estimation algorithms. Using daily physical activity data, we demonstrate that an automatic sensor localization algorithm decreases the estimation error of the MET calculation by a factor of 2.3 compared to the case without sensor localization. Furthermore, our sensor localization algorithm achieves an accuracy of 90.8% in detecting on-body locations of wearable sensors. The integration of sensor localization and MET estimation achieves an accuracy of 80% in calculating the MET values of daily physical activities.

An Energy-Efficient Computational Model for Uncertainty Management in Dynamically Changing Networked Wearables

The utility of wearables is currently limited to lab experiments and controlled environments mainly because computational algorithms embedded in wearables fail to produce accurate measurements in uncontrolled, dynamically changing, and potentially harsh environments. With the exponentially growing adoption of these systems in human-centered Internet-of-Things (IoT) applications, development of resource-efficient solutions to enhance the accuracy of this systems remains a considerable research challenge. In this paper, we introduce an energy-efficient framework for uncertainty management of networked wearables. The core components of our framework are anomaly screening units for detecting anomalies that require handling, thus resulting in one order of magnitude less energy consumption compared to the conventional frameworks. Furthermore, our screening approach achieves 98.3% accuracy in detecting anomalies based on real data collected with wearable motion sensors.

A Reliable and Reconfigurable Signal Processing Framework for Estimation of Metabolic Equivalent of Task in Wearable Sensors

Wearable motion sensors are widely used to estimate metabolic equivalent of task (MET) values associated with physical activities. However, one major obstacle in widespread adoption of current wearables is that any changes in configuration of the network requires new data collection and re-training of the underlying signal processing algorithms. For any wearable-based MET estimation framework to be considered a viable platform, it needs to be reconfigurable, reliable, and power-efficient. In this paper, we aim to address the issues of sensor misplacement, power efficiency, and new sensor addition and propose a reliable and reconfigurable MET estimation framework. We introduce a power-aware sensor localization approach that allows users to wear the sensors on different body locations without need for adhering to a specific installation protocol. Furthermore, we propose a novel transductive transfer learning approach, which gives end-users the ability to add new sensors to the network without need for collecting new training data. This is accomplished by transferring the knowledge of already trained sensors to the untrained sensors in real-time. Our experiments demonstrate that our sensor localization algorithm achieves an accuracy of 90.8% in detecting location of the wearable sensors. The integrated model of sensor localization and MET calculation achieves an R2 of 0.8 in estimating MET values using a regression-based model. Furthermore, our transfer learning algorithm improves the R2 value of MET estimation up to 60%.

Smart-Cuff: A wearable bio-sensing platform with activity-sensitive information quality assessment for monitoring ankle edema

Leg swelling produced by retention of fluid in leg tissues is known as peripheral edema, which is regarded as a symptom for various systematic diseases such as heart or kidney failure. In current clinical practice, edema is manually assessed by clinical experts. Such an assessment can often be inaccurate and unreliable especially if it is made by different operators at different times. Despite the importance of monitoring edema for the purpose of evaluating the course of disease or the effect of treatment, quantifying peripheral edema in a continuous and accurate fashion has remained a challenge. In this paper, we propose a wearable real-time platform (namely, Smart-Cuff), which integrates advanced technologies in sensing, computation, and signal processing and machine learning for continuous and real-time edema monitoring in remote and in-home settings. Given that peripheral edema is highly dependent on various contextual attributes such as body posture, we present an activity-sensitive approach to discard erroneous or contextually invalid sensor data in order to meet the requirements of both energy efficiency and quality of information. Examination of our hardware prototype demonstrates the effectiveness of the proposed force-sensitive resistor-based edema sensor (with an R2 of 0.97 for our regression model) as well as the activity monitoring mechanism (over 99% accuracy) that provide the means to perform reliable data sanity check on ankle circumference measurements in a continuous manner.

Explicit estimation-error-probability computation and sensor design for flag Hidden Markov Models

Hidden Markov Models (HMM) are used in a number of sensor networking applications. These applications often require performance evaluation and sensor design for HMM estimation algorithms. This article approaches the performance evaluation and design problems from a structural perspective. Specifically, for a special class of flag HMMs (where sensors accurately flag a subset of states), explicit formulae are derived for the average error probability of the maximum-likelihood estimate. These formulae are used to optimally place sensors, and to gain an understanding of the relationship between the HMMs structure and estimation error. Three examples, including a real-world case study on monitoring the elderly in a smart home, are presented.

Transfer learning algorithms for autonomous reconfiguration of wearable systems

Wearables have emerged as a revolutionary technology in many application domains including healthcare and fitness. Machine learning algorithms, which form the core intelligence of wearables, traditionally deduce a computational model from a set of training examples to detect events of interest (e.g. activity type). However, in the dynamic environment in which wearables typically operate in, the accuracy of a computational model drops whenever changes in configuration of the system (such as device type and sensor orientation) occur. Therefore, there is a need to develop systems which can adapt to the new configuration autonomously. In this paper, using transfer learning as an organizing principle, we develop several algorithms for data mapping. The data mapping algorithms employ effective signal similarity methods and are used to adapt the system to the new configuration. We demonstrate the efficacy of the data mapping algorithms using a publicly available dataset on human activity recognition.

Digital memory notebook: experimental evaluation of motivational reward strategies

Prompting technology can help individuals with cognitive impairments complete independent activities of daily living (IADL). Although the prompt delivery is an effective way to remind an adult to record a completed activity, this potential benefit may not be sufficient to motivate the adult to comply with the prompt on a consistent basis. In this work we extend activity-aware prompting techniques to utilize alternative reward structures. Our reward mechanism will allow adults to observe game progress as a result of their decisions to comply with the prompts. In our study with volunteer participants, the activity-aware reward-based prompting method increased the compliance rate compared to activity-aware prompting without rewarding the adults.