Accuracy Comparison of CNN, LSTM, and Transformer for Activity Recognition Using IMU and Visual Markers

Abstract

Human activity recognition (HAR) has applications ranging from security to healthcare. Typically these systems are composed of data acquisition and activity recognition models. In this work, we compared the accuracy of two acquisition systems: Inertial Measurement Units (IMUs) vs Movement Analysis Systems (MAS). We trained models to recognize arm exercises using state-of-the-art deep learning architectures and compared their accuracy. MAS uses a camera array and reflective markers. IMU uses accelerometers, gyroscopes, and magnetometers. Sensors of both systems were attached to different locations of the upper limb. We captured and annotated 3 datasets, each one using both systems simultaneously. For activity recognition, we trained 8 architectures, each one with different operations and layers configurations. The best architectures were a combination of CNN, LSTM, and Transformer achieving test accuracy from 89% to 99% on average. We evaluated how feature selection reduced the sensors required. We found IMU and MAS data were able to distinguish correctly the arm exercises. CNN layers at the beginning produced better accuracy on challenging datasets. IMU had advantages over other aquisition systems for activity recognition. We analyzed the relations between models accuracy, signal waveforms, signals correlation, sampling rate, exercise duration, and window size. Finally, we proposed the use of a single IMU located at the wrist and a variable-size window extraction.

Otago Exercises Monitoring for Older Adults by a Single IMU and Hierarchical Machine Learning Models

“…The scenarios using a large window (long time series) for OEP recognition could take advantage of CNN-BiLSTM. Similar results were also reported by a previous study, reporting that CNN-LSTM outperformed CNN, LSTM, and Transformer for IMU-based activity recognition [18]. However, deep learning models require more training examples than classical models.…”

“…This architecture has been validated for many datasets with better recognition performance than both CNN and LSTM [19], [20]. It also outperformed Transformers in recent research for HAR [18]. Based on this architecture and Bi-directional-LSTM (BiLSTM) layers, the CNN-BiLSTM was further developed [21], [22].…”

“…For HAR systems, Long Short-Term Memory (LSTM) as a type of RNN units has also been proven efficient for long time series [16]. Besides, Transformers have also been explored for HAR using attention mechanism [17], [18]. The mechanism searches for correlation between features by mapping a query and a set of keys, which makes it efficient for long-time series recognition.…”

A CNN Model for Physical Activity Recognition and Energy Expenditure Estimation from an Eyeglass-Mounted Wearable Sensor

“…Nevertheless, the superior performance of deep learning, such as convolutional neural networks (CNNs), long short-term memory (LSTM), and attention mechanisms, makes them attractive for PAR and EE estimation applications [33,34]. In [35], CNN and LSTM with self-attention mechanisms improved PAR performance, and a similar experiment in [36] enhanced arm exercise activity recognition. CNN-LSTM models showed improved performance and were used by other recent researchers [37,38].…”