Association between upper limb movements during drumming and cognition in older adults with cognitive impairment and dementia at a nursing home: a pilot study.

Background: Despite the association between motor dysfunction and dementia, quantitative assessment of dementia-related specific motor dysfunction in patients with severe dementia is difficult. Thus, this study aimed to develop a new method to measure upper limb motor function in people with dementia. Methods: We examined the relationship between dementia severity and dementia-related specific motor dysfunction using the Mini-Mental State Examination (MMSE), a dementia screening test. Participants comprised 16 nursing home residents with a mean age of 86 years and MMSE score of 14.56 (range, 1-23) Points. Participants were seated in a circle and instructed to play a drum that was placed in their lap using mallets (drumsticks) in their dominant hand. Acceleration and gyroscopic sensors were attached to their wrists to collect data on arm movements while drumming. Upper limb motor characteristics were confirmed by recording acceleration and arm movement during drumming and analyzing the correlation with handgrip strength. Results: Handgrip strength was correlated with arm elevation angle during drumming. The arm elevation angle displayed a significant regression equation with the MMSE score and showed the best regression equation along with handgrip strength (adjusted R2 = 0.6035, p = 0.0009). Conclusion: We developed a new method using drums to measure upper limb motor function in people with dementia. We also verified that the average arm elevation angle during drumming could predict cognitive dysfunction. This system may be used to monitor people with dementia in a simple and safe way.

Recent Progress on Triboelectric Nanogenerators for Vibration Energy Harvesting and Vibration Sensing.

The triboelectric nanogenerator (TENG) is a recent technology that reforms kinetic energy generation and motion sensing. A TENG comes with variety of structures and mechanisms that make it suitable for wide range of applications and working conditions. Since mechanical vibrations are abundant source of energy in the surrounding environment, the development of a TENG for vibration energy harvesting and vibration measurements has attracted a huge attention and great research interest through the past two decades. Due to the high output voltage and high-power density of a TENG, it can be used as a sustainable power supply for small electronics, smart devices, and wireless sensors. In addition, it can work as a vibration sensor with high sensitivity. This article reviews the recent progress in the development of a TENG for vibration energy harvesting and vibration measurements. Systems of only a TENG or a hybrid TENG with other transduction technologies, such as piezoelectric and electromagnetic, can be utilized for vibrations scavenging. Vibration measurement can be done by measuring either vibration displacement or vibration acceleration. Each can provide full information about the vibration amplitude and frequency. Some TENG vibration-sensing architectures may also be used for energy harvesting due to their large output power. Numerous applications can rely on TENG vibration sensors such as machine condition monitoring, structure health monitoring, and the Internet of things (IoT).

Natural Frequency Response Evaluation for RC Beams Affected by Steel Corrosion Using Acceleration Sensors.

This paper presented a laboratory investigation for analyzing the natural frequency response of reinforced concrete (RC) beams affected by steel corrosion. The electrochemical acceleration technique induced the corroded RC beams until the predetermined value of the steel corrosion ratio was achieved. Then, the natural frequency responses of the corroded beams were tested utilizing piezoelectric acceleration sensors. The damage states of the corroded beams were assessed through the measurement of crack parameters and the equivalent elastic modulus of the beams, which aims to clarify the fundamental characteristics of the dynamic response for the corroded RC beam with the increased steel corrosion ratio. The results revealed that steel corrosion reduces the bending stiffness of the RC beams and, thus, reduces the modal frequency. The variation of natural frequency can identify the corrosion damage even if no surface cracking of the RC beam, and the second-order frequency should be more indicative of the damage scenario. The degradations of stiffness and the natural frequency were estimated in this study by the free vibration equation of a simply supported beam, and a prediction method for the RC beam's residual service life was established. This study supports the use of variations in natural frequency as one diagnostic indicator to evaluate the health of RC bridge structures.

[Human activity recognition based on the inertial information and convolutional neural network].

With the rapid improvement of the perception and computing capacity of mobile devices such as smart phones, human activity recognition using mobile devices as the carrier has been a new research hot-spot. The inertial information collected by the acceleration sensor in the smart mobile device is used for human activity recognition. Compared with the common computer vision recognition, it has the following advantages: convenience, low cost, and better reflection of the essence of human motion. Based on the WISDM data set collected by smart phones, the inertial navigation information and the deep learning algorithm-convolutional neural network (CNN) were adopted to build a human activity recognition model in this paper. The K nearest neighbor algorithm (KNN) and the random forest algorithm were compared with the CNN network in the recognition accuracy to evaluate the performance of the CNN network. The classification accuracy of CNN model reached 92.73%, which was much higher than KNN and random forest. Experimental results show that the CNN algorithm model can achieve more accurate human activity recognition and has broad application prospects in predicting and promoting human health.

Human activity recognition based on the inertial information and convolutional neural network / 生物医学工程学杂志

With the rapid improvement of the perception and computing capacity of mobile devices such as smart phones, human activity recognition using mobile devices as the carrier has been a new research hot-spot. The inertial information collected by the acceleration sensor in the smart mobile device is used for human activity recognition. Compared with the common computer vision recognition, it has the following advantages: convenience, low cost, and better reflection of the essence of human motion. Based on the WISDM data set collected by smart phones, the inertial navigation information and the deep learning algorithm-convolutional neural network (CNN) were adopted to build a human activity recognition model in this paper. The K nearest neighbor algorithm (KNN) and the random forest algorithm were compared with the CNN network in the recognition accuracy to evaluate the performance of the CNN network. The classification accuracy of CNN model reached 92.73%, which was much higher than KNN and random forest. Experimental results show that the CNN algorithm model can achieve more accurate human activity recognition and has broad application prospects in predicting and promoting human health.

[A Domestic Diagnosis System for Early Restless Legs Syndrome Based on Deep Learning].

Restless legs syndrome,as a common sleep disorder,has nowadays long been diagnosed by self-rating scale and polysomnography.In this paper,a domestic diagnosis system for early restless legs syndrome based on deep learning is proposed,which is suitable for early patients with unstable symptoms in routine diagnosis.The hardware system is installed in the bed.And the non-contact sleeping dynamic signal acquisition is realized based on the acceleration sensors.The software system uses deep learning to classify and recognize the signals.A Fully Connected Feedforward Network based on Keras framework is constructed to recognize seven kinds of activities during sleeping.The accuracy of comprehensive classification is 97.83%.Based on former results,the periodic limb movement index and awakening index were evaluated to make the diagnosis of restless legs syndrome.

Communication Rate Increase in Drill Strings of Oil and Gas Wells Using Multiple Actuators.

Wireless data communication and telemetry during drilling deep oil and gas wells are important enablers for safe and timely drilling operations. The transmission of information through drill strings and pipes using sound waves is a useful and practical approach. However, given the limited available bandwidth, transmission rates are typically smaller than what is needed. In this paper, a new method and system are proposed to increase the transmission rate over the same bandwidth, by deploying more than one actuator. Upon using multiple actuators, several data streams can be transmitted simultaneously. This increases the data rate without the need for additional bandwidth. The experimental results of a testbed with two actuators are presented, where the transmission rate is doubled with no bandwidth increase. A strain sensor receiver and accelerometer receivers are used to separate and demodulate the two data streams. It is demonstrated that it is possible to recover the data in the new faster system benefiting from two actuators, while having about the same bit error probability performance as a one-actuator system. Various combinations of strain and acceleration sensors are considered at the receive side. Due to some properties of strain channels (e.g., smaller delay spreads and their less-frequency-selective behavior) presented in this paper, it appears that a strain sensor receiver and an accelerometer receiver together can offer a good performance when separating and demodulating the two actuators' data in the testbed. Overall, the experimental results from the proposed system suggest that upon using more than one actuator, it is feasible to increase the data rate over the limited bandwidth of pipes and drill strings.

A Domestic Diagnosis System for Early Restless Legs Syndrome Based on Deep Learning / 中国医疗器械杂志

Restless legs syndrome,as a common sleep disorder,has nowadays long been diagnosed by self-rating scale and polysomnography.In this paper,a domestic diagnosis system for early restless legs syndrome based on deep learning is proposed,which is suitable for early patients with unstable symptoms in routine diagnosis.The hardware system is installed in the bed.And the non-contact sleeping dynamic signal acquisition is realized based on the acceleration sensors.The software system uses deep learning to classify and recognize the signals.A Fully Connected Feedforward Network based on Keras framework is constructed to recognize seven kinds of activities during sleeping.The accuracy of comprehensive classification is 97.83%.Based on former results,the periodic limb movement index and awakening index were evaluated to make the diagnosis of restless legs syndrome.

Acceleration sensors in abdominal wall position as a non-invasive approach to detect early breathing alterations induced by intolerance of increased airway resistance.

BACKGROUND: Early detection of respiratory overload is crucial to mechanically ventilated patients, especially during phases of spontaneous breathing. Although a diversity of methods and indices has been established, there is no highly specific approach to predict respiratory failure. This study aimed to evaluate acceleration sensors in abdominal and thoracic wall positions to detect alterations in breathing excursions in a setting of gradual increasing airway resistance. METHODS: Twenty-nine healthy volunteers were committed to a standardized protocol of a two-minutes step-down spontaneous breathing on a 5 mm, 4 mm and then 3 mm orally placed endotracheal tube. Accelerator sensors in thoracic and abdominal wall position monitored breathing excursions. 15 participants passed the breathing protocol ("completed" group), 14 individuals cancelled the protocol due to subjective intolerance to the increasing airway resistance ("abandoned" group). RESULTS: Gradual increased respiratory workload led to a significant decrease of acceleration in abdominal wall position in the "abandoned" group compared to the "completed" group (p < 0.001), while these gradual accelerating changes were not observed in thoracic wall position (p = 0.484). Thoracic acceleration sensors did not detect any time- and group-specific changes (p = 0.746). CONCLUSIONS: The abdominal wall position of the acceleration sensors may be a non-invasive, economical and practical approach to detect early breathing alterations prior to respiratory failure. TRIAL REGISTRATION: EK 309-15; by the Ethics Committee of the Faculty of Medicine, RWTH Aachen, Aachen, Germany. Retrospectively registered 28th of December 2015.

Wearable Sensor Data Classification for Human Activity Recognition Based on an Iterative Learning Framework.

The design of multiple human activity recognition applications in areas such as healthcare, sports and safety relies on wearable sensor technologies. However, when making decisions based on the data acquired by such sensors in practical situations, several factors related to sensor data alignment, data losses, and noise, among other experimental constraints, deteriorate data quality and model accuracy. To tackle these issues, this paper presents a data-driven iterative learning framework to classify human locomotion activities such as walk, stand, lie, and sit, extracted from the Opportunity dataset. Data acquired by twelve 3-axial acceleration sensors and seven inertial measurement units are initially de-noised using a two-stage consecutive filtering approach combining a band-pass Finite Impulse Response (FIR) and a wavelet filter. A series of statistical parameters are extracted from the kinematical features, including the principal components and singular value decomposition of roll, pitch, yaw and the norm of the axial components. The novel interactive learning procedure is then applied in order to minimize the number of samples required to classify human locomotion activities. Only those samples that are most distant from the centroids of data clusters, according to a measure presented in the paper, are selected as candidates for the training dataset. The newly built dataset is then used to train an SVM multi-class classifier. The latter will produce the lowest prediction error. The proposed learning framework ensures a high level of robustness to variations in the quality of input data, while only using a much lower number of training samples and therefore a much shorter training time, which is an important consideration given the large size of the dataset.