In this article, the self-supported power conditioning circuits are studied for a footstep energy harvester, which consists of a monolithic multilayer piezoelectric stack with a force amplification frame to extract electricity from human walking locomotion. Based on the synchronized switch harvesting on inductance (SSHI) technology, the power conditioning circuits are designed to optimize the power flow from the piezoelectric stack to the energy storage device under real-time human walking excitation instead of a simple sine waveform input, as reported in most literatures. The unique properties of human walking locomotion and multilayer piezoelectric stack both impose complications for circuit design. Three common interface circuits, for example, standard energy harvesting circuit, series-SSHI, and parallel-SSHI, are compared in terms of their output power to find the best candidate for the real-time-footstep energy harvester. Experimental results show that the use of parallel-SSHI circuit interface produces 74% more power than the standard energy harvesting counterpart, while the use of series-SSHI circuit demonstrates a similar performance in comparison to the standard energy harvesting interface. The reasons for such a huge efficiency improvement using the parallel-SSHI interface are detailed in this article.

Gait monitoring and analysis has attracted accumulative attention from the health technology community, with a particular high demand on wearable devices enabling long-term real-time daily gait monitoring. However, the vast majority of existing commercial and emerging technologies use significant electric power, need multiple sensor nodes, and are expensive and complicated to wear. This paper proposes a customized low power, low-cost monitoring insole (MONI) with minimized sensor nodes for real time daily gait monitoring and analysis in an accurate and reliable manner. With one accelerometer positioned at the heel and one at the first metatarsal, key gait features can be extracted and recorded in real-time. Selected gait information will be constructed as personalized daily gait traces to provide a low-power wearable solution for observing user’s gait evolution, which is critical for disease detection at early stages, such as Parkinson’s disease. To minimize the power consumption while efficiently collecting scattered gait information throughout the day, a working-mode management (WMM) algorithm, and an auto activity recognition (AAR) algorithm are developed to intelligently sample the right amount of true walking gait data in real-time. MONI is tested and validated both in the lab (with an accuracy of over 88.74% and a strong correlation intra class correlations (ICC) over 0.83). A pilot study is conducted in the field, suggesting less than 2-mW daily power consumption for 5500 steps of walking for an office worker.

Harnessing footstep energy allows for a unique method of power generation from a largely untapped resource, with applications such as instantly charging mobile and wearable electronic devices. This letter presents an easily integrated heel charger to efficiently convert kinematic walking energy into electricity. The heel charger uses a multilayered levered piezoelectric (L-Pie) mechanism associated with a nonlinear mechanical-synchronized switching on inductor circuit (M-SSHI) interface. It tactfully switches on when the foot contacts the ground, and switches off when the foot is lifted. This design takes full advantage of the user's weight, and amplifies footstep displacement by utilizing the lever mechanism to gain maximum deformation of multilayer piezoelectric patches. The experimental results show that the fabricated two-level multilayer L-pie has a top performance of 13.60 mW of AC RMS, a mechanical to AC power conversion efficiency of 7.87%, and a DC RMS power of 6.13 mW, an AC to DC power conversion efficiency of 45.07%, under a harmonic excitation of 2.3 Hz (mimicking fast walking speed: 6.2 km/h for men, 5.55 km/h for women). With an optimal load of 210 kΩ, the two-level L-pie using the M-SSHI has an improvement of 206.45% in DC RMS power compared to a standard 4-diode bridge energy harvesting circuit. The L-pie design works more efficiently under large force excitation, even with small displacement, which makes this technology optimal for footstep energy harvesting.

Foot gestures play an important role in human machine interface and also indicate the lower-limb motor functionalities. Lately, machine learning models have been integrated with different kinds of wearable devices for foot gesture recognition. However, most existing wearable devices lack of robustness and facileness and existing machine learning models suffer from inconsistent recognition accuracy due to limited and unbalanced training data sets, and therefore do not work well for new users. To address this, a customized convolutional neural network model (CNN) architecture is designed and strategies of achieving personalized models are presented, considering potential fairness sensitive attributes of a diverse training group, including age, gender, body mass index (defined by height and weight), shoe size and health conditions. The proposed model to be embodied into a flexible smart insole with only two accelerometers aims to recognize four types of foot gestures - toe tapping, heel tapping, foot kicking and foot stepping in a sitting posture at high accuracies. After the hyperparameters are well-tuned, an average recognition accuracy reaches 92% in training sets, 88.05% in validation sets, 85.27% in test sets and 83.09% in 5-fold cross-validation. The evaluation results from both in-group and leave-one-subject-out show our unified model is easily transformable for individual uses by further fine-tuning the key hyperparameters.

As Parkinson’s disease (PD) progresses, patients experience difficult times to distinguish their medication effective (ON) and non-effective (OFF) periods, due to subjective judgements on observations of motor symptoms and self-recording in diaries. Thus, a simple-to-use tool to objectively identify and record ON and OFF in daily life is in great need. This study proposes a method to distinguish medication ON and OFF for PD patients through 4 repetitive foot motions – toe tapping, heel tapping, toe pivoting and heel pivoting by wearing a pair of monitoring insoles (MONI). After collecting data from 16 patients and 9 age-matched healthy controls, we extract over 1200 features from all repetitive foot motions and provides comprehensive analytical results of feature importance with Analysis on Variance (ANOVA). Then, we use Random Forest Classifier to first identify foot motions through motion recognition and then extract features to distinguish ON and OFF. Through evaluation on 2-month long data collected from 5 individual patients, we identify motions performed mistakenly by one patient during an experimental session and show 85% ON/OFF distinguishment accuracy for the rest 20 sessions, compared to 60% accuracy from a baseline model referencing to observation judgement in clinical tests. The proposed method and experimental procedures can be transformed into an assistant tool, including an application (APP) and MONI, as a potential solution to help patients with Parkinson’s disease understand and objectively record their medication effectiveness in daily life.

Postoperative complications pose substantial health risks to children who undergo surgery, yet timely detection of complications after discharge is challenging due to reliance on subjective symptom reports from children and caregivers. Alternatively, wearable devices can provide objective health measurements for continuous recovery monitoring, potentially enabling earlier complication detection in the hospital or community. This study examined biorhythm-based metrics (circadian and ultradian rhythms, derived from the daily activity and heart rate patterns recorded by a consumer wearable) and their relationship to postoperative recovery in children with and without complications. Wearables were given to 103 children for 21 days immediately after appendectomy, and biorhythm metrics were extracted from per-minute data. A machine-learned model using these metrics retrospectively predicted postoperative complications up to 3 days before formal diagnosis with 91% sensitivity and 74% specificity. Our findings suggest that wearable-derived biorhythms offer a promising, unobtrusive method for evaluating postoperative recovery. This approach has broad clinical implications for pediatric health monitoring across various care settings.