Without expert coaching, inexperienced exercisers performing core exercises, such as squats, are subject to an increased risk of spinal or knee injuries. Although it is theoretically possible to measure the kinematics of body segments and classify exercise forms with wearable sensors and algorithms, the current implementations are not sufficiently accurate. In this study, the squat posture classification performance of deep learning was compared to that of conventional machine learning. Additionally, the location for the optimal placement of sensors was determined. Accelerometer and gyroscope data were collected from 39 healthy participants using five inertial measurement units (IMUs) attached to the left thigh, right thigh, left calf, right calf, and lumbar region. Each participant performed six repetitions of an acceptable squat and five incorrect forms of squats that are typically observed in inexperienced exercisers. The accuracies of squat posture classification obtained using conventional machine learning and deep learning were compared. Each result was obtained using one IMU or a combination of two or five IMUs. When employing five IMUs, the accuracy of squat posture classification using conventional machine learning was 75.4%, whereas the accuracy using deep learning was 91.7%. When employing two IMUs, the highest accuracy (88.7%) was obtained using deep learning for a combination of IMUs on the right thigh and right calf. The single IMU yielded the best results on the right thigh, with an accuracy of 58.7% for conventional machine learning and 80.9% for deep learning. Overall, the results obtained using deep learning were superior to those obtained using conventional machine learning for both single and multiple IMUs. With regard to the convenience of use in self-fitness, the most feasible strategy was to utilize a single IMU on the right thigh.

Although many smartphone application (app) programs provide education and guidance for basic life support, they do not commonly provide feedback on the chest compression depth (CCD) and rate. The validation of its accuracy has not been reported to date. This study was a feasibility assessment of use of the smartphone as a CCD feedback device. In this study, we proposed the concept of a new real-time CCD estimation algorithm using a smartphone and evaluated the accuracy of the algorithm. Using the double integration of the acceleration signal, which was obtained from the accelerometer in the smartphone, we estimated the CCD in real time. Based on its periodicity, we removed the bias error from the accelerometer. To evaluate this instrument's accuracy, we used a potentiometer as the reference depth measurement. The evaluation experiments included three levels of CCD (insufficient, adequate, and excessive) and four types of grasping orientations with various compression directions. We used the difference between the reference measurement and the estimated depth as the error. The error was calculated for each compression. When chest compressions were performed with adequate depth for the patient who was lying on a flat floor, the mean (standard deviation) of the errors was 1.43 (1.00) mm. When the patient was lying on an oblique floor, the mean (standard deviation) of the errors was 3.13 (1.88) mm. The error of the CCD estimation was tolerable for the algorithm to be used in the smartphone-based CCD feedback app to compress more than 51 mm, which is the 2010 American Heart Association guideline.

A virtual blind cane system for indoor application, including a camera, a line laser and an inertial measurement unit (IMU), is proposed in this paper. Working as a blind cane, the proposed system helps a blind person find the type of obstacle and the distance to it. The distance from the user to the obstacle is estimated by extracting the laser coordinate points on the obstacle, as well as tracking the system pointing angle. The paper provides a simple method to classify the obstacle’s type by analyzing the laser intersection histogram. Real experimental results are presented to show the validity and accuracy of the proposed system.

Stooping is a posture which is described as an involuntary forward bending of the thoracolumbar spine. Conventionally, the stooped posture (SP) in Parkinson’s disease patients is measured in static or limited movement conditions using a radiological or optoelectronic system. In the dynamic condition with long movement distance, there was no effective method in preference to the empirical assessment from doctors. In this research, we proposed a practical method for estimating the SP with a high accuracy where accelerometers can be mounted on the neck or upper back as a wearable sensor. The experiments with simulated subjects showed a high correlation of 0.96 and 0.99 between the estimated SP angle and the reference angles for neck and back sensor position, respectively. The maximum absolute error (0.9 and 1.5 degrees) indicated that the system can be used, not only in clinical assessment as a measurement, but also in daily life as a corrector.