Using time series data obtained from accelerometer and gyroscope sensors on an iPhone 6s, the authors address the topic of human activity and attribute detection. The collection contains time series data from 24 subjects who completed six activities in 15 trials. The aim is to appropriately identify the six activities using machine learning techniques. The usage of a convolutional neural network (CNN) for the categorization of human activity and attribute identification data obtained from accelerometer and gyroscope sensors on an iPhone 6s is proposed in this research. The collection contains time series data from 24 subjects who completed six activities in 15 trials. The study begins by pre-processing the data by transforming the folders into class labels and plotting the time series data. The time-series data is made up of multivariate data from both the accelerometer and gyroscope sensors, totaling 12 characteristics. The accelerometer sums up two acceleration vectors, gravity, and user acceleration, which may be distinguished using core motion tracking technology.
TopIntroduction
The Human Activity and Attribute Recognition: Phone Accelerometer and Gyroscope dataset, which comprises time-series data generated by accelerometer and gyroscope sensors, is one such dataset that has been widely utilised for this purpose. The dataset contains information from 24 people of various ages, genders, weights, and heights who participated in 15 trials of six distinct activities (walking downstairs, upstairs, sitting, standing, running, and walking) in the same location and conditions. For each trial, the dataset includes a multivariate time-series with 12 features such as attitude, gravity, userAcceleration, and rotationRate. The accelerometer calculates the total of two acceleration vectors: user acceleration and gravity, while the gyroscope tracks the device's orientation. In these 12 feature, the Acceleration has 3 features, X, Y and Z. As a result, the dataset records both the physical activity and the device's orientation during the activity. Human activity detection has gotten a lot of interest in recent years, both to the increasing availability of sensor data and the necessity for automated analysis of human behaviour in a variety of scenarios. The utilisation of time series data from wearable sensors, which can provide a plethora of information on human movement and behaviour, has been an important field of research. Such information has been widely used in applications, including health monitoring and rehabilitation, sports performance analysis, and security.
Human Activity Recognition Based on Sensors Data
The classification of activities conducted by individuals using sensors attached to their bodies is a component of HAR. The goal is to precisely detect individuals' actions, which can have a variety of applications in domains such as sports training, healthcare, and behaviour monitoring. The availability of low-cost sensors, their non-intrusive nature, and the potential to collect data in real-world scenarios have all contributed to the increased interest in HAR employing wearable sensors. Human activity recognition is one such application where time series data analysis is crucial. Based on data received from sensors, it recognises various human behaviours such as walking, jogging, sitting, standing, and others. Human activity recognition offers a wide range of applications, including healthcare monitoring, sports performance analysis, and geriatric care.
Human Activity Recognition (HAR) and Smartphones and Wearable Devices
HAR has significant consequences in a variety of industries, including healthcare, sports, and entertainment. In healthcare, for example, it can be used to monitor the everyday elderly people activities or patients with chronic conditions in order to spot any changes in their behaviour or prospective health problems. It can be used in sports to analyse athlete performance and give personalised training strategies. It can be used to build interactive games or immersive virtual reality experiences in the entertainment industry. The use of huge datasets that record a varied range of activities conducted by a number of persons is required for the creation of accurate and trustworthy HAR models. The use of smartphones and wearable devices in recent years has permitted the capture of large-scale sensor data in a more naturalistic setting. For example, (Gupta et al., 2022) discuss the use of artificial intelligence in medical data analysis
The analysis of sensor data supplied by accelerometer and gyroscope sensors in cell phones is the topic of this paper. We intend to investigate the feasibility of using such data to recognise human actions and infer personal qualities such as gender or personality. We anticipate that our research will add to the increasing corpus of HAR research and provide useful insights into the possibilities of IoT sensor data for human activity detection. As an example, (Talukdar et al. 2022) and (Veeraiah et al. 2022) explore the use of machine learning and IoT for suspicious activity detection and enhancing the capabilities of the metaverse, respectively.