Enhanced Human Activity Recognition (HAR) with IMU Sensors in Smartphones: Insights from Machine Learning Models

Human activity recognition is a very active research on pervasive computing and mobile health application. Many human activity systems based on inertial measurement unit (IMU) sensor data were proposed in the past few years. These systems mainly use IMU sensor placed on he torso and limbs to collect data and utilize supervised machine learning algorithms on sensor data. One main issue of these systems is that wearing multiple on-body IMU sensors may bring inconvenience to users' daily life. The other issue of these exiting methods is that an activity recognition model that is trained on a specific subject does not work well when being applied to predict another subject's activities since IMU activity data always carry information that is specific to the human subject who conducts the activities. In our work, inspired by the principle of domain adaption, we proposed a new deep-learning activity recognition model based on an adversarial network which can remove the subject-specific information within the IMU activity data and extract subject-independent features shared by the data collected on different subjects. We also for the first time use data collected from insole based IMU sensors on 8 participants for 5 common activities to build a new real world human activity dataset which can minimize the inconvenience for users to wear. We conducted experiments with our new real-world dataset. Results show that our subject independent activity recognition model outperforms state-of-art supervised learning techniques and eliminates the effects of individual differences between subjects successfully. The average recognition accuracy under the leave-one-out (L1O) condition achieves 99.0% which is higher than the performance of traditional human activity recognition system based on CNNs.

The study of human activity recognition (HAR) holds significant importance within wearable technology and ubiquitous computing, driven by the increasing ubiquity of inertial measurement unit (IMU) sensors embedded in devices like smartphones, smartwatches, and fitness trackers. The effective classification and recognition of human actions are crucial for various applications, including health monitoring, fitness tracking, and personalized user experiences. This study comprehensively examines the advancements in HAR by applying machine learning (ML) methodologies to data collected from IMU sensors. We explore seven powerful ML algorithms that have been pivotal in transforming raw sensor data into actionable insights for activity classification. These algorithms include decision trees, random forests, support vector machines (SVM), k-nearest neighbors (KNN), artificial neural networks (ANN), convolutional neural networks (CNN), and long short-term memory networks (LSTM). Each algorithm is assessed based on its ability to accurately process and classify various human activities, highlighting their strengths and limitations in different scenarios. Moreover, the study delves into the critical role of evaluation metrics and the confusion matrix in validating the performance of these ML models. Metrics such as accuracy, precision, recall, F1 score, and specificity are examined to provide a holistic view of the model's efficacy. The confusion matrix is emphasized as a tool for understanding the true positive, false positive, true negative, and false negative rates, offering insights into the practical performance of the models in realworld applications. Through this detailed investigation, we aim to shed light on the current state of HAR and the potential future directions for research and development in this dynamic field.

Human activity recognition (HAR) plays a pivotal role in various domains, including healthcare, sports, robotics, and security. With the growing popularity of wearable devices, particularly Inertial Measurement Units (IMUs) and Ambient sensors, researchers and engineers have sought to take advantage of these advances to accurately and efficiently detect and classify human activities. This research paper presents an advanced methodology for human activity and localization recognition, utilizing smartphone IMU, Ambient, GPS, and Audio sensor data from two public benchmark datasets: the Opportunity dataset and the Extrasensory dataset. The Opportunity dataset was collected from 12 subjects participating in a range of daily activities, and it captures data from various body-worn and object-associated sensors. The Extrasensory dataset features data from 60 participants, including thousands of data samples from smartphone and smartwatch sensors, labeled with a wide array of human activities. Our study incorporates novel feature extraction techniques for signal, GPS, and audio sensor data. Specifically, for localization, GPS, audio, and IMU sensors are utilized, while IMU and Ambient sensors are employed for locomotion activity recognition. To achieve accurate activity classification, state-of-the-art deep learning techniques, such as convolutional neural networks (CNN) and long short-term memory (LSTM), have been explored. For indoor/outdoor activities, CNNs are applied, while LSTMs are utilized for locomotion activity recognition. The proposed system has been evaluated using the k-fold cross-validation method, achieving accuracy rates of 97% and 89% for locomotion activity over the Opportunity and Extrasensory datasets, respectively, and 96% for indoor/outdoor activity over the Extrasensory dataset. These results highlight the efficiency of our methodology in accurately detecting various human activities, showing its potential for real-world applications. Moreover, the research paper introduces a hybrid system that combines machine learning and deep learning features, enhancing activity recognition performance by leveraging the strengths of both approaches.

While inertial measurement units (IMU) have become an integral part of sports performance analysis, upper body-mounted IMUs have been found to exhibit poor reliability in measuring lower-limb loading. In racket sports, IMUs have been placed in a number of positions on the upper body, lower body and racket in a research setting. A potential limitation to the concurrent use of multiple IMUs is that coaches may be reluctant to allow their athletes to wear the units during training and competition due to concerns that the units would interfere with athlete movement. This study seeks to understand the perceptions of racket sports coaches towards the use of IMUs in training and competition. A total of 58 racket sport coaches responded to a survey on the use of IMUs during training and competition. Based on the responses, 96.6% (56 out of 58) of coaches indicated that they would allow their athletes to wear IMUs in training, while 65.5% (38 out of 58) would allow their athletes to wear IMUs during competition. For use in training, 9 of the 14 suggested IMU placements received significant positive responses. However, none of the suggested IMU placements received significant positive responses for use during competition and 11 of the 14 received significant negative responses. This suggests that while coaches understand the benefits of collecting data from IMUs during competition. Despite this, for use in training, a number of upper and lower body-mounted IMUs placements have the potential to be part of regular monitoring in racket sports.

Human activity recognition (HAR) is the study of the identification of specific human movement and action based on images, accelerometer data and inertia measurement unit (IMU) sensors. In the sensor based HAR application, most of the researchers used many IMU sensors to get an accurate HAR classification. The use of many IMU sensors not only limits the deployment phase but also increase the difficulty and discomfort for users. As reported in the literature, the original model used 19 sensor data consisting of accelerometers and IMU sensors. The imbalanced class distribution is another challenge to the recognition of human activity in real-life. This is a real-life scenario, and the classifier may predict some of the imbalanced classes with very high accuracy. When a model is trained using an imbalanced dataset, it can degrade model’s performance. In this paper, two approaches, namely resampling and multiclass focal loss, were used to address the imbalanced dataset. The resampling method was used to reconstruct the imbalanced class distribution of the IMU sensor dataset prior to model development and learning using the cross-entropy loss function. A deep ConvLSTM network with a minimal number of IMU sensor data was used to develop the upper-body HAR model. On the other hand, the multiclass focal loss function was used in the HAR model and classified minority classes without the need to resample the imbalanced dataset. Based on the experiments results, the developed HAR model using a cross-entropy loss function and reconstructed dataset achieved a good performance of 0.91 in the model accuracy and F1-score. The HAR model with a multiclass focal loss function and imbalanced dataset has a slightly lower model accuracy and F1-score in both 1% difference from the resampling method. In conclusion, the upper body HAR model using a minimal number of IMU sensors and proper handling of imbalanced class distribution by the resampling method is useful for the assessment of home-based rehabilitation involving activities of daily living.

Mitochondrial recessive ataxia syndrome (MIRAS) is a heritable disease, relatively common in Finland. Among other things, patients suffer from ataxia, a movement disorder with difficulties in coordination. To date, no treatment is known for the disease, but medication and therapy can lessen the symptoms, provided that the progression of symptoms is closely monitored to adjust the treatment according to the individual needs. This necessary evaluation is a manual, subjective process.We report about our efforts to explore quantifiable characteristics that could be used to monitor the disease progression objectively using electromyography (EMG) as well as inertial measurement unit (IMU) sensors. In particular, in a study with eight participants, including a patient, we have collected muscle activation as well as IMU data during several tasks. The study found some characteristics that might qualify as indicators of ataxia, such as high-frequency electrical activity (EA) components and similarity of repetitions. We further suggest the use of IMU and machine learning to improve the objective monitoring of the disease’s progression.

In recent years, human activity monitoring and recognition have gained importance in providing valuable information to improve the quality of life. A lack of activity can cause health problems including falling, depression, and decreased mobility. Continuous activity monitoring can be useful to prevent progressive health problems. With this purpose, this study presents a wireless smart insole with four force-sensitive resistors (FSRs) that monitor foot contact states during activities for both indoor and outdoor use. The designed insole is a compact solution and provides walking comfort with a slim and flexible structure. Moreover, the inertial measurement unit (IMU) sensors designed in our previous study were used to collect 3-axis accelerometer and 3-axis gyroscope outputs. Smart insoles were located in the shoe sole for both right and left feet, and two IMU sensors were attached to the thigh area of each leg. The sensor outputs were collected and recorded from forty healthy volunteers for eight different gait-based activities including walking uphill and descending stairs. The obtained datasets were separated into three categories; foot contact states, the combination of acceleration and gyroscope outputs, and a set of all sensor outputs. The dataset for each category was separately fed into deep learning algorithms, namely, convolutional long–short-term memory neural networks. The performance of each neural network for each category type was examined. The results show that the neural network using only foot contact states presents 90.1% accuracy and provides better performance than the combination of acceleration and gyroscope datasets for activity recognition. Moreover, the neural network presents the best results with 93.4% accuracy using a combination of all the data compared with the other two categories.

Inertial measurement units (IMU) constitute a light and cost-effective alternative to gold-standard measurement systems in the assessment of running temporal variables. IMU data collected on 20 runners running at different speeds (80, 90, 100, 110 and 120% of preferred running speed) and treadmill inclination (±2, ±5, and ±8%) were used here to predict the following temporal variables: stride frequency, duty factor, and two indices of running variability such as the detrended fluctuation analysis alpha (DFA-α) and the Higuchi’s D (HG-D). Three different estimation methodologies were compared: 1) a gold-standard optoelectronic device (which provided the reference values), 2) IMU placed on the runner’s feet, 3) a single IMU on the runner’s thorax used in conjunction with a machine learning algorithm with a short 2-second or a long 120-second window as input. A two-way ANOVA was used to test the presence of significant (p<0.05) differences due to the running condition or to the estimation methodology. The findings of this study suggest that using both IMU configurations for estimating stride frequency can be effective and comparable to the gold-standard. Additionally, the results indicate that the use of a single IMU on the thorax with a machine learning algorithm can lead to more accurate estimates of duty factor than the strategy of the IMU on the feet. However, caution should be exercised when using these techniques to measure running variability indices. Estimating DFA-α from a short 2-second time window was possible only in level running but not in downhill running and it could not accurately estimate HG-D across all running conditions. By taking a long 120-second window a machine learning algorithm could improve the accuracy in the estimation of DFA-α in all running conditions. By taking these factors into account, researchers and practitioners can make informed decisions about the use of IMU technology in measuring running biomechanics.

Inertial measurement units (IMU) constitute a light and cost-effective alternative to gold-standard measurement systems in the assessment of running temporal variables. IMU data collected on 20 runners running at different speeds (80, 90, 100, 110 and 120% of preferred running speed) and treadmill inclination (±2, ±5, and ±8%) were used here to predict the following temporal variables: stride frequency, duty factor, and two indices of running variability such as the detrended fluctuation analysis alpha (DFA-α) and the Higuchi's D (HG-D). Three different estimation methodologies were compared: 1) a gold-standard optoelectronic device (which provided the reference values), 2) IMU placed on the runner's feet, 3) a single IMU on the runner's thorax used in conjunction with a machine learning algorithm with a short 2-second or a long 120-second window as input. A two-way ANOVA was used to test the presence of significant (p<0.05) differences due to the running condition or to the estimation methodology. The findings of this study suggest that using both IMU configurations for estimating stride frequency can be effective and comparable to the gold-standard. Additionally, the results indicate that the use of a single IMU on the thorax with a machine learning algorithm can lead to more accurate estimates of duty factor than the strategy of the IMU on the feet. However, caution should be exercised when using these techniques to measure running variability indices. Estimating DFA-α from a short 2-second time window was possible only in level running but not in downhill running and it could not accurately estimate HG-D across all running conditions. By taking a long 120-second window a machine learning algorithm could improve the accuracy in the estimation of DFA-α in all running conditions. By taking these factors into account, researchers and practitioners can make informed decisions about the use of IMU technology in measuring running biomechanics.

Human Activity Recognition (HAR) is an active research field because of its versatility towards various application areas such as healthcare and lifecare. In this study, a novel HAR system is proposed based on an autoencoder for denoising and Recurrent Neural Network (RNN) for classification with a single Inertial Measurement Unit (IMU) located on a dominant wrist. A Variational Autoencoder (VAE) is built to denoise IMU signals which improves HAR by Android Deep RNN. Evaluating our VAE and Android Deep RNN HAR system is done in two ways. First, the system is tested on a PC using discrete epochs of activities of daily living. Our results show that VAE improves Signal-to-Noise Ratio (SNR) of the IMU signal from 8.78 to 17.26 dB. In turn, HAR improves from 89.29% to 95.11% in F1-score and from 90.38% to 95.47% in accuracy. Secondly, the system is tested on an Android device (i.e., smartphone) using continuous activity signals. This is done by transferring the PC HAR system to an Android HAR App (i.e., Android Deep RNN). We have achieved 86.13% and 95.09% in accuracy without and with VAE respectively. Our results demonstrate that HAR can be achieved in real-time on a standalone smart device with a single IMU for lifelogging services.

This work presents an on-device machine learning model with the ability to identify different mobility gestures called human activity recognition (HAR), which includes running, walking, squatting, jumping, and others. The data is collected through an Arduino Nano 33 BLE Sense board with a sampling rate of 119 Hz, which is embedded with an Inertial Measurement Unit (IMU) sensor. The same board is used as a microcontroller to identify human gestures by developing an end-to-end edge computing application. A deep neural network model is trained and then compressed for deployment on the board to create a self-contained, embedded device capable of identifying the type of gesture performed. Three deep learning models, namely Multi-Layer Perceptron (MLP), Convolutional Neural Network – Long Short Term Memory (CNN-LSTM) & CNN-Gated Recurrent Unit (CNN-GRU), are evaluated in the identification of the mobility gestures. The observed accuracy of the models is 96%, 97.1% and 97.8%, respectively, MLP, CNN-LSTM & CNN-GRU across the different gesture categories. The study shows the utility of embedded devices with deep neural network-based models, which can provide low cost, minimal power usage, and meet data privacy requirements in HAR.

&amp;lt;p&amp;gt;The use of Inertial Measurement Units (IMUs) in geomorphological studies has exploded during the last decade. Scientists are deploying IMUs in a range of settings: from single grain flume experiments to full scale landslide motions and from capturing rock falls to measuring flows in glacial environments.&amp;lt;/p&amp;gt;&amp;lt;p&amp;gt;The vast majority of these experiments deploy sensing units that are partly customised for each application. However, there are limits to the level of IMU customisation geomorphologists can do as they rarely have access to bottom-up sensor assembly and production lines. Commercial IMUs and IMU components are built and calibrated for very different uses than the monitoring of dynamic sediment transport regimes, such as integration into electronic devices, wearables or Internet of Things applications.&amp;lt;/p&amp;gt;&amp;lt;p&amp;gt;Deploying commercial IMUs outside their nominal operational range has two main implications, the first being methodological. As the sensor is partly a &amp;quot;black box&amp;quot;, we are obliged to do extensive testing in a trial-and-error manner and think deeply about the underlying physics of IMUs. If such difficulties are not acknowledged the results become difficult to interpret in the context of sediment movement.&amp;lt;/p&amp;gt;&amp;lt;p&amp;gt;The second implication concerns standardisation. The more our community uses commercial sensors and analytical tools, the more apparent becomes the need for open-source pre-processing and processing workflows that are fully validated and universally available to ensure comparability of published results.&amp;lt;/p&amp;gt;&amp;lt;p&amp;gt;This presentation aspires to contribute to this open debate about IMU sensors in geomorphology. The focus will be on the sensing requirements for grain motion detection, force capture and tracking by IMUs in the context of sediment transport. The presented calculations will use results published before the emergence of IMUs in geomorphology for a range of environments (fluvial, coastal, aeolian and glacial).&amp;lt;/p&amp;gt;&amp;lt;p&amp;gt;The above requirements capture will be accompanied by a meta-analysis of published IMU data in geomorphic applications which will be classified according to the exact type of sensor (accelerometer, full IMU, GPS (or equivalent)-aided IMU) and the sensors' specs (mainly sensing range and frequency).&amp;lt;/p&amp;gt;&amp;lt;p&amp;gt;Finally, this presentation will explore the case study of using a commercially available IMU for the capture of fluvial sediment interactions. The deployed IMU will be subjected to a series of simple physical experiments (e.g., drop tests) and then deployed to a flume setting designed to model grain-grain and grain-substrate collisions. The novelty here is the use of an independent very high-speed camera (1&amp;amp;#956;s exposure frame rate) to monitor the sensor during calibration, which allows for the coherent propagation of uncertainty for all the experiments. All the results are presented within a processing workflow based on free, open-source R libraries.&amp;lt;/p&amp;gt;

PurposeThis paper aims to deal with the human activity recognition using human gait pattern. The paper has considered the experiment results of seven different activities: normal walk, jogging, walking on toe, walking on heel, upstairs, downstairs and sit-ups.Design/methodology/approachIn this current research, the data is collected for different activities using tri-axial inertial measurement unit (IMU) sensor enabled with three-axis accelerometer to capture the spatial data, three-axis gyroscopes to capture the orientation around axis and 3° magnetometer. It was wirelessly connected to the receiver. The IMU sensor is placed at the centre of mass position of each subject. The data is collected for 30 subjects including 11 females and 19 males of different age groups between 10 and 45 years. The captured data is pre-processed using different filters and cubic spline techniques. After processing, the data are labelled into seven activities. For data acquisition, a Python-based GUI has been designed to analyse and display the processed data. The data is further classified using four different deep learning model: deep neural network, bidirectional-long short-term memory (BLSTM), convolution neural network (CNN) and CNN-LSTM. The model classification accuracy of different classifiers is reported to be 58%, 84%, 86% and 90%.FindingsThe activities recognition using gait was obtained in an open environment. All data is collected using an IMU sensor enabled with gyroscope, accelerometer and magnetometer in both offline and real-time activity recognition using gait. Both sensors showed their usefulness in empirical capability to capture a precised data during all seven activities. The inverse kinematics algorithm is solved to calculate the joint angle from spatial data for all six joints hip, knee, ankle of left and right leg.Practical implicationsThis work helps to recognize the walking activity using gait pattern analysis. Further, it helps to understand the different joint angle patterns during different activities. A system is designed for real-time analysis of human walking activity using gait. A standalone real-time system has been designed and realized for analysis of these seven different activities.Originality/valueThe data is collected through IMU sensors for seven activities with equal timestamp without noise and data loss using wirelessly. The setup is useful for the data collection in an open environment outside the laboratory environment for activity recognition. The paper also presents the analysis of all seven different activity trajectories patterns.

The proper execution of the sprint start is crucial in determining the performance during a sprint race. In this respect, when moving from the crouch to the upright position, trunk kinematics is a key element. The purpose of this study was to validate the use of a trunk-mounted inertial measurement unit (IMU) in estimating the trunk inclination and angular velocity in the sagittal plane during the sprint start. In-laboratory sprint starts were performed by five sprinters. The local acceleration and angular velocity components provided by the IMU were processed using an adaptive Kalman filter. The accuracy of the IMU inclination estimate and its consistency with trunk inclination were assessed using reference stereophotogrammetric measurements. A Bland-Altman analysis, carried out using parameters (minimum, maximum, and mean values) extracted from the time histories of the estimated variables, and curve similarity analysis (correlation coefficient > 0.99, root mean square difference < 7 deg) indicated the agreement between reference and IMU estimates, opening a promising scenario for an accurate in-field use of IMUs for sprint start performance assessment.

Human activity recognition (HAR) has recently gained popularity due to its applications in healthcare, surveillance, human‐robot interaction, and various other fields. Deep learning (DL)‐based models have been successfully applied to the raw data captured through inertial measurement unit (IMU) sensors to recognize multiple human activities. Despite the success of DL‐based models in human activity recognition, feature extraction remains challenging due to class imbalance and noisy data. Additionally, selecting optimal hyperparameter values for DL models is essential since they affect model performance. The hyperparameter values of some of the existing DL‐based HAR models are chosen randomly or through the trial‐and‐error method. The random selection of these significant hyperparameters may be suitable for some applications, but sometimes it may worsen the model's performance in others. Hence, to address the above‐mentioned issues, this research aims to develop an optimized DL model capable of recognizing various human activities captured through IMU sensors. The proposed DL‐based HAR model combines convolutional neural network (CNN) layers and bidirectional long short‐term memory (Bi‐LSTM) units to simultaneously extract spatial and temporal sequence features from raw sensor data. The Rao‐3 metaheuristic optimization algorithm has been adopted to identify the ideal hyperparameter values for the proposed DL model in order to enhance its recognition performance. The proposed DL model's performance is validated on PAMAP2, UCI‐HAR, and MHEALTH datasets and achieved 94.91%, 97.16%, and 99.25% accuracies, respectively. The results reveal that the proposed DL model performs better than the existing state‐of‐the‐art (SoTA) models.