Efficient FPGA Implementation of Multilayer Perceptron for Real-Time Human Activity Classification

Abstract

The smartphone-based human activity recognition (HAR) systems are not capable to deliver high-end performance for challenging applications. We propose a dedicated hardware-based HAR system for smart military wearables, which uses a multilayer perceptron (MLP) algorithm to perform activity classification. To achieve the flexible and efficient hardware design, the inherent MLP architecture with parallel computation is implemented on FPGA. The system performance has been evaluated using the UCI human activity dataset with 7767 feature samples of 20 subjects. The three combinations of a dataset are trained, validated, and tested on ten different MLP models with distinct topologies. The MLP design with the 7-6-5 topology is finalized from the classification accuracy and cross entropy performance. The five versions of the final MLP design (7-6-5) with different data precision are implemented on FPGA. The analysis shows that the MLP designed with 16-bit fixed-point data precision is the most efficient MLP implementation in the context of classification accuracy, resource utilization, and power consumption. The proposed MLP design requires only 270 ns for classification and consumes 120 mW of power. The recognition accuracy and hardware results performance achieved are better than many of the recently reported works. INDEX TERMS Human activity recognition, multilayer perceptron, smart military wearables, classifier hardware implementation, field programmable gate array.

FANN-on-MCU: An Open-Source Toolkit for Energy-Efficient Neural Network Inference at the Edge of the Internet of Things

“…Finally, for very small networks, such as the ones used in applications B and C, the runtime is far below the millisecond range. If the application scenario requires only very few classifications per cluster activation, then the IBEX core is the most energy-efficient one, with a consumption of 2.9 µJ and 0.15 µJ, respectively for applications B and C. Comparing to the work in [46] for application C, the IBEX core is 13.5× faster in computation time and 434× more energy efficient than a parallel FPGA implementation. However, if continuous classification is required, which is the case for the vast majority of the IoT applications, then the parallel execution, once again, outperforms in terms of speed and energy efficiency.…”

“…MLPs have been successfully used in a wide range of application scenarios, such as disease detection [45], activity recognition [46], and brain-machine interface [47]. Many studies identified MLPs to be the best or one of the best algorithms to solve tasks in the IoT domain using wearable devices [48]- [51].…”

“…The authors in [46] proposed an FPGA implementation of MLPs with parallel computation to classify human activity in real-time. The dataset is acquired by a 3-axial accelerometer wore on the waist and classified into five activity classes.…”

Embedded Artificial Neural Network FPGA Controlled Cart

“…Gaikwad et al [29] proposed a hardware implemented FPGA for military equipment that uses an MLP algorithm to perform classification tasks. Parallel MLP computation was implemented to reach enhanced hardware design.…”

A Model-Based Design Floating-Point Accumulator. Case of Study: FPGA Implementation of a Support Vector Machine Kernel Function

“…The development of wearable devices leads to implementation of ML algorithms directly on board [18,19], allowing for the reduction of the amount of data to be transmitted, and with consistent advantages in terms of power consumption and system usability [14]. To address the issues related to the need for platforms with good computing capacity, instead of general-purpose processors, dedicated hardware architectures such as field programmable gate arrays (FPGAs) can be selected for the implementation of the algorithms [20][21][22][23]. This allows for the control of the resources needed for the task and to optimize the system for performance or physical size, depending on the use case.…”