IMU-Based Fitness Activity Recognition Using CNNs for Time Series Classification

Philipp Niklas Müller,Alexander Josef Müller,Philipp Achenbach,Stefan Göbel

DOI: https://doi.org/10.3390/s24030742

IF: 3.9

2024-01-24

Sensors

Abstract:Mobile fitness applications provide the opportunity to show users real-time feedback on their current fitness activity. For such applications, it is essential to accurately track the user's current fitness activity using available mobile sensors, such as inertial measurement units (IMUs). Convolutional neural networks (CNNs) have been shown to produce strong results in different time series classification tasks, including the recognition of daily living activities. However, fitness activities can present unique challenges to the human activity recognition task (HAR), including greater similarity between individual activities and fewer available data for model training. In this paper, we evaluate the applicability of CNNs to the fitness activity recognition task (FAR) using IMU data and determine the impact of input data size and sensor count on performance. For this purpose, we adapted three existing CNN architectures to the FAR task and designed a fourth CNN variant, which we call the scaling fully convolutional network (Scaling-FCN). We designed a preprocessing pipeline and recorded a running exercise data set with 20 participants, in which we evaluated the respective recognition performances of the four networks, comparing them with three traditional machine learning (ML) methods commonly used in HAR. Although CNN architectures achieve at least 94% test accuracy in all scenarios, two traditional ML architectures surpass them in the default scenario, with support vector machines (SVMs) achieving 99.00 ± 0.34% test accuracy. The removal of all sensors except one foot sensor reduced the performance of traditional ML architectures but improved the performance of CNN architectures on our data set, with our Scaling-FCN reaching the highest accuracy of 99.86 ± 0.11% on the test set. Our results suggest that CNNs are generally well suited for fitness activity recognition, and noticeable performance improvements can be achieved if sensors are dropped selectively, although traditional ML architectures can still compete with or even surpass CNNs when favorable input data are utilized.

engineering, electrical & electronic,chemistry, analytical,instruments & instrumentation

What problem does this paper attempt to address?

### What problems does this paper attempt to solve?

This paper aims to evaluate the applicability of convolutional neural networks (CNNs) in fitness activity recognition tasks (FAR) based on inertial measurement units (IMUs) and compare their performance with traditional machine - learning methods. Specifically, the main research objectives include:

1. **Adapting existing CNN architectures**: Adjust three existing CNN architectures to fit the fitness activity recognition task.
2. **Designing a new CNN architecture**: Design a new CNN architecture - Scaling - FCN - to deal with the problem of varying execution speeds in fitness activities.
3. **Dataset creation**: Create a new public dataset containing IMU data of 20 participants in seven different running exercises.
4. **Pre - processing pipeline**: Develop a pipeline for fitness activity data recording and pre - processing.
5. **Performance evaluation**: Evaluate the performance of four CNN architectures and three traditional machine - learning methods under different input data parameters, especially the impact of the number of sensors and the size of input data on performance.

### Research background

Mobile fitness applications provide users with real - time feedback, helping them get better guidance in non - stationary fitness activities. To achieve this goal, it is crucial to accurately identify the user's current fitness activity. IMUs, as wearable sensors, can be used to track users' fitness activities in real - time in mobile applications. Although CNNs have performed well in time - series classification tasks, in fitness activity recognition, they face some unique challenges, such as limited training data, high similarity between different activities, and limited computing resources on mobile devices.

### Main contributions

1. **Proposing a new CNN architecture**: Introduce the Scaling - FCN architecture designed specifically for sensor - based fitness activity recognition in fixed - time windows.
2. **Creating a public dataset**: Provide a new public dataset containing IMU data of 20 participants in seven different running exercises.
3. **Developing a pre - processing pipeline**: Design a set of pipelines for recording and pre - processing fitness activity data from multiple sensors.
4. **Detailed performance analysis**: Conduct a detailed comparative analysis of the performance of Scaling - FCN with the other three existing CNN architectures and three traditional machine - learning methods on the running exercise dataset, focusing on the impact of different input data parameters.

Through these efforts, the researchers hope to prove the potential of CNNs in fitness activity recognition and explore how to improve model performance by selectively reducing the number of sensors.