Electrical and Electronic Engineering Research Collection

We consider the downlink of a cell-free massive multiple-input multiple-output (MIMO) system where large number of access points (APs) simultaneously serve a group of users. Two fundamental problems are of interest, namely (i) to maximize the total spectral efficiency (SE), and (ii) to maximize the minimum SE of all users. As the considered problems are non-convex, existing solutions rely on successive convex approximation to find a sub-optimal solution. The known methods use off-the-shelf convex solvers, which basically implement an interior-point algorithm, to solve the derived convex problems. The main issue of such methods is that their complexity does not scale favorably with the problem size, limiting previous studies to cell-free massive MIMO of moderate scales. Thus the potential of cell-free massive MIMO has not been fully understood. To address this issue, we propose an accelerated projected gradient method to solve the considered problems. Particularly, the proposed solution is found in closed-form expressions and only requires the first order information of the objective, rather than the Hessian matrix as in known solutions, and thus is much more memory efficient. Numerical results demonstrate that our proposed solution achieves far less run-time, compared to other second-order methods.

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 4×4 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.

Wearable inertial sensors offer the possibility to monitor sleeping position and respiration rate during sleep, enabling a comfortable and low-cost method to remotely monitor patients. Novel methods to estimate respiration rate and position during sleep using accelerometer data are presented, with algorithm performance examined for two sensor locations, and accelerometer-derived respiration rate compared across sleeping positions. Eleven participants (9 male; aged: 47.82±14.14 years; BMI 30.9±5.27 kg/m 2 ; AHI 5.77±4.18) undergoing a scheduled clinical polysomnography (PSG) wore a tri-axial accelerometer on their chest and upper abdomen. PSG cannula flow and position data were used as benchmark data for respiration rate (breaths per minute, bpm) and position. Sleeping position was classified using logistic regression, with features derived from filtered acceleration and orientation. Accelerometer-derived respiration rate was estimated for 30 s epochs using an adaptive peak detection algorithm which combined filtered acceleration and orientation data to identify individual breaths. Sensor-derived and PSG respiration rates were then compared. Mean absolute error (MAE) in respiration rate did not vary between sensor locations (abdomen: 1.67±0.37 bpm; chest: 1.89±0.53 bpm; p=0.52), while reduced MAE was observed when participants lay on their side (1.58±0.54 bpm) compared to supine (2.43±0.95 bpm), p<; 0.01. MAE was less than 2 bpm for 83.6% of all 30 s windows across all subjects. The position classifier distinguished supine and left/right with a ROC AUC of 0.87, and between left and right with a ROC AUC of 0.94. The proposed methods may enable a low-cost solution for in-home, long term sleeping posture and respiration monitoring.

We develop a novel technique to accelerate minorization-maximization (MM) procedure for the non-orthogonal multiple access (NOMA) weighted sum rate maximization problem. Specifically, we exploit the Lipschitz continuity of the gradient of the objective function to adaptively update the MM algorithm. With fewer additional analysis variables and low complexity second-order cone program (SOCP) to solve in each iteration of the MM algorithm, the proposed approach converges quickly at a small computational cost. By numerical simulation results, our algorithm is shown to greatly outperform known solutions in terms of achieved sum rates and computational complexity.

Wearable accelerometers can be used to quantify movement during swimming, enabling objective performance analysis. This study examined arm acceleration during front crawl swimming, and investigated how accelerometer-derived features change with lap times. Thirteen participants swam eight 50m laps using front crawl with a tri-axial accelerometer attached to each upper arm. Data were segmented into individual laps; lap times estimated and individual strokes extracted. Stroke times, root mean squared (RMS) acceleration, RMS jerk and spectral edge frequencies (SEF) were calculated for each stroke. Movement symmetry was assessed as the ratio of the minimum to maximum feature value for left and right arms. A regularized multivariate regression model was developed to estimate lap time using a subset of the accelerometer-derived features. Mean lap time was 56.99±11.99s. Fifteen of the 42 derived features were significantly correlated with lap time. The regression model included 5 features (stroke count, mean SEF of the X and Z axes, stroke count symmetry, and the coefficient of variation of stroke time symmetry) and estimated 50m lap time with a correlation coefficient of 0.86, and a cross-validated RMS error of 6.38s. The accelerometer-derived features and developed regression model may provide a useful tool to quantitatively evaluate swimming performance.