Advances in mobile technology have led to the emergence of the “smartphone”, a new class of device with more advanced connectivity features that have quickly made it a constant presence in our lives. Smartphones are equipped with comparatively advanced computing capabilities, a global positioning system (GPS) receivers, and sensing capabilities (i.e., an inertial measurement unit (IMU) and more recently magnetometer and barometer) which can be found in wearable ambulatory monitors (WAMs). As a result, algorithms initially developed for WAMs that “count” steps (i.e., pedometers); gauge physical activity levels; indirectly estimate energy expenditure and monitor human movement can be utilised on the smartphone. These algorithms may enable clinicians to “close the loop” by prescribing timely interventions to improve or maintain wellbeing in populations who are at risk of falling or suffer from a chronic disease whose progression is linked to a reduction in movement and mobility. The ubiquitous nature of smartphone technology makes it the ideal platform from which human movement can be remotely monitored without the expense of purchasing, and inconvenience of using, a dedicated WAM. In this paper, an overview of the sensors that can be found in the smartphone are presented, followed by a summary of the developments in this field with an emphasis on the evolution of algorithms used to classify human movement. The limitations identified in the literature will be discussed, as well as suggestions about future research directions.

A geometrically intuitive quaternion-based complementary filter has been developed, which estimates the attitude of a smartphone using signals obtained from the device's internal magnetic and inertial measurement unit. This method introduces two correction factors μ _a , which governs the rate at which the accelerometer corrects the pitch/roll angle, and μ _m , which controls the rate at which the yaw angle is corrected by the magnetometer. The smartphone's attitude can be determined by combining information from the tri-axial gyroscope (measures the rate at which the device rotates), the tri-axial accelerometer (measures the acceleration due to gravity in the absence of body movement), and the triaxial magnetometer (can be used to determine the direction of geomagnetic north). In order to validate the algorithm, a nonferrous rectangular plate with retroreflective markers affixed at each corner was attached to the smartphone enabling its attitude to be tracked by an optical motion capture system. The accuracy of the algorithm (φ _RMSE = 3.37°, θ _RMSE = 1.84°, ψ _RMSE = 4.83°) was assessed as ten subjects performed five normal daily activities while the smartphone was loosely fixed to the upper thigh to simulate being carried in a pants pocket. The optimal tuning parameters (μ _a = 0.003 and μ _m = 0.001) for the algorithm across all the body movements were also identified.

Automatic recognition of human activity is useful as a means of estimating energy expenditure and has potential for use in fall detection and prediction. The emergence of the smartphone as a ubiquitous device presents an opportunity to utilize its embedded sensors, computational power and data connectivity as a platform for continuous health monitoring. In the study described herein, 37 older people (83.9 ± 3.4 years) performed a series of activities of daily living (ADLs) while a smartphone (containing a triaxial accelerometer, triaxial gyroscope and barometric pressure sensor) was placed in the front pocket of their trousers. These results are compared to a similar trial conducted previously in which 20 young people (21.9 ± 1.65 years) were asked to perform the same ADLs using the same smartphone (again in the front pocket of their trousers).In each trial, the participants were asked to perform several activities (standing, sitting, lying, walking on level ground, up and down staircases, and riding an elevator up and down) in a free-living environment. During each acquisition session, the internal sensor signals were recorded and subsequently used to develop activity classifiers based on a decision tree algorithm that classified ADL in epochs of ~1.25 s. When training and testing with the younger cohort, using a leave-one-out cross validation procedure, a total classification sensitivity of 80.9% ± 9.57% ([Formula: see text] = 0.75 ± 0.12) was obtained. Retraining and testing on the older cohort, again using cross validation, gives a comparable total class sensitivity of 82.0% ± 8.88% ([Formula: see text] =0.74 ± 0.12).When trained with the younger group and tested on the older group, a total class sensitivity of 69.2% ± 24.8% (95% confidence interval [69.6%, 70.6%]) and [Formula: see text] = 0.60 ± 0.27 (95% confidence interval [0.58, 0.59]) was obtained. When trained on the older group and tested on the younger group, a total class sensitivity of 80.5% ± 6.80% (95% confidence interval [79.0%, 80.6%]) and [Formula: see text] = 0.74 ± 0.08 (95% confidence interval [0.73, 0.75]) was obtained.An instance of the decision tree classifier developed was implemented on the smartphone as a software application. It was capable of performing real-time activity classification for a period of 17 h on a single battery charge, illustrating that smartphone technology provides a viable platform on which to perform long-term activity monitoring.

Goal: This paper presents an algorithm for accurately estimating pelvis, thigh, and shank kinematics during walking using only three wearable inertial sensors. Methods: The algorithm makes novel use of a constrained Kalman filter (CKF). The algorithm iterates through the prediction (kinematic equation), measurement (pelvis position pseudo-measurements, zero velocity update, flat-floor assumption, and covariance limiter), and constraint update (formulation of hinged knee joints and ball-and-socket hip joints). Results: Evaluation of the algorithm using an optical motion capture-based sensor-to-segment calibration on nine participants (7 men and 2 women, weight 63.0 6.8 kg, height 1.70 0.06 m, age 24.6 3.9 years old), with no known gait or lower body biomechanical abnormalities, who walked within a $4 \times 4 m^2$ capture area shows that it can track motion relative to the mid-pelvis origin with mean position and orientation (no bias) root-mean-square error (RMSE) of 5.21 1.3 cm and 16.1 3.2, respectively. The sagittal knee and hip joint angle RMSEs (no bias) were 10.0 2.9 and 9.9 3.2, respectively, while the corresponding correlation coefficient (CC) values were 0.87 0.08 and 0.74 0.12. Conclusion: The CKF-based algorithm was able to track the 3D pose of the pelvis, thigh, and shanks using only three inertial sensors worn on the pelvis and shanks. Significance: Due to the Kalman-filter-based algorithm's low computation cost and the relative convenience of using only three wearable sensors, gait parameters can be computed in real-time and remotely for long-term gait monitoring. Furthermore, the system can be used to inform real-time gait assistive devices.

A geometrically intuitive quaternion-based complementary attitude and heading reference system (CAHRS) proposed in our previous work estimated the attitude of a magnetic and inertial measurement unit (MIMU). The method used two correction factors, $\mu _{_{a}}$ that determined the rate at which the accelerometer corrected the inclination angle, and $\mu _{_{m}}$ that governed the rate at which the magnetometer corrected the yaw angle. Improvements to the filter have been made by embedding each correction factor within an error-state Kalman filter (KF), enabling the correction rates to behave adaptively. The revised filter only estimates the error in two variables, thus remaining computationally efficient (65 addition, 88 subtraction, and 214 multiplication operations) compared with established algorithms in the literature for attitude estimation that utilize a KF or extended KF. The accuracy of the attitude estimated (i.e., the pitch, roll, and yaw angle errors $\theta _{\text {RMSE}_{_{}}}^{\circ }$ , $\phi _{\text {RMSE}}^{\circ }$ , and $\psi _{\text {RMSE}}^{\circ }$ ) by the adaptive error-state KF was compared with the CAHRS algorithm and a cascaded KF that is representative of state-of-the-art methods. Each algorithm was assessed using a publicly available data set in which the attitude of a foot-worn magnetic and inertial measurement units was recorded by a motion capture system while participants walked and ran around a room for one or three minutes ( $\phi _{\text {RMSE}}^{\circ } = 2.08^\circ $ , $\theta _{\text {RMSE}}^{\circ } = 1.98^\circ $ , and $\psi _{\text {RMSE}}^{\circ }= 5.25^\circ $ ).