Low Complexity Binarized 2D-CNN Classifier for Wearable Edge AI Devices

Wearable Artificial Intelligence-of-Things (AIoT) devices exhibit the need to be resource and energy-efficient. In this paper, we introduced a quantized multilayer perceptron (qMLP) for converting ECG signals to binary image, which can be combined with binary convolutional neural network (bCNN) for classification. We deploy our model into a low-power and low-resource field programmable gate array (FPGA) fabric. The model requires 5.8x lesser multiply and accumulate (MAC) operations than known wearable CNN models. Our model also achieves a classification accuracy of 98.5%, sensitivity of 85.4%, specificity of 99.5%, precision of 93.3%, and F1-score of 89.2%, along with dynamic power dissipation of 34.9 mu W.

Automated summary

In this paper, a quantized multilayer perceptron (qMLP) is introduced to convert ECG signals to binary images, which can be classified using a binary convolutional neural network (bCNN). The model is deployed on a low-power and low-resource FPGA fabric, requiring fewer multiply and accumulate operations compared to known wearable CNN models. The model achieves high classification accuracy and other performance metrics, along with low power dissipation.