Two billion people are affected by hemoglobin (Hgb) related diseases. Usual clinical assessments of Hgb are conducted by analyzing venipuncture-obtained blood samples in laboratories. A non-invasive, cheap, point-of-care and accurate Hgb test is needed everywhere. Our group has developed a noninvasive Hgb measurement system using 10-second Smartphone videos of the index fingertips. Custom hardware sets were used to illuminate the fingers. We tested four lighting conditions with wavelengths in the near-infrared spectrum suggested by the absorption properties of two primary components of bloodoxygenated Hgb and plasma. We found a strong linear correlation between our measured and laboratory-measured Hgb levels in 167 patients with a mean absolute percentage error (MAPE) of 5%. In our initial analysis, critical tasks were performed manually. Now, using the same data, we have automated or modified all the steps. For all, male, and female subjects we found a MAPE of 6.43%, 5.34%, and 4.85 and mean squared error (MSE) of 0.84, 0.5, and 0.49 respectively. The new analyses however, have suggested inexplicable inconsistencies in our results, which we attribute to laboratory measurement errors reflected in a non-normative distribution of Hgb levels in our studied patients, as well as excess noise in the specific signals we measured in the videos. Based on these encouraging results, and the promise of greater accuracy with our revised hardware and software tools, we now propose a rigorous validation study to demonstrate that this approach to hemoglobin measurement is appropriate for general clinical application.

BACKGROUND Human Activity Recognition is a widely researched topic. Researchers have recently put significant effort into detecting and measuring different types of physical activity. Due to their low cost, sensors are one of the most used components to recognize activities performed by humans. With the recent emergence of machine learning techniques, real-time activity monitoring is possible with raw sensor data. Estimation of energy expenditure while performing activities is also essential in healthcare research. Detection of performed physical and social activities and the exertion related to these activities are beneficial for patients with limited physical ability. Elderly patients with Chronic Fatigue Syndrome / Myalgic Encephalomyelitis (CFS/ME) become fatigued with little effort and later suffer from malaise and poor and unrefreshing sleep. They need to know their daily energy expenditure with their activities at any time to avoid fatigue. Therefore, a comprehensive study of physical and social activity detection and analysis of the existing methods of energy expenditure calculation in the healthcare domain is required. OBJECTIVE The first objective of this paper is to explore existing activity recognition methods and find out the best practices in processing sensor data, machine learning techniques and their challenges, and the approaches to detect social exertion. Secondly, we want to compare existing energy expenditure measurement techniques mentioned in the literature. METHODS We searched for articles related to physical and social activity recognition and methods of calculating energy expenditure. We compared the sensors used in the research to find the best possible combination and configuration for different activity types. We also explored the literature’s pre-processing, feature extraction, and machine-learning techniques applied to raw sensor data. RESULTS Accelerometers are the most used sensor to get important information about leg and hand activities. Gyroscopes and gravity sensors are the following choices of sensors combined with the accelerometer. Sensors are set at a frequency range of 20-50 Hz. Smartphones are the most popular choice for data collection. Many sensor-based healthcare researchers use smartphones as a point-of-care solution to collect data and provide feedback based on the data. Wrist-worn sensor devices provide better context than smartphones for hand-performed activities. The microphone of smartphones and smartwatches can detect social exertion and stress levels from contextual information about the surrounding. Objective measures are the best way to get a proper and unbiased energy expenditure. Metabolic Equivalent of Task (MET) is the most popular unit for intensity measurement. CONCLUSIONS This literature comprehensively reviews recent sensor-based research to detect and measure physical activities. There are very few systems that detect and measure physical activities by using sensors. With the outgrowing number of patients with limited capability of performing day-to-day life activities, we need improved methods to develop such a system.