FPGA Implementation for Odor Identification with Depthwise Separable Convolutional Neural Network

Abstract

The integrated electronic nose (e-nose) design, which integrates sensor arrays and recognition algorithms, has been widely used in different fields. However, the current integrated e-nose system usually suffers from the problem of low accuracy with simple algorithm structure and slow speed with complex algorithm structure. In this article, we propose a method for implementing a deep neural network for odor identification in a small-scale Field-Programmable Gate Array (FPGA). First, a lightweight odor identification with depthwise separable convolutional neural network (OI-DSCNN) is proposed to reduce parameters and accelerate hardware implementation performance. Next, the OI-DSCNN is implemented in a Zynq-7020 SoC chip based on the quantization method, namely, the saturation-flooring KL divergence scheme (SF-KL). The OI-DSCNN was conducted on the Chinese herbal medicine dataset, and simulation experiments and hardware implementation validate its effectiveness. These findings shed light on quick and accurate odor identification in the FPGA.

COVID-DSNet: A novel deep convolutional neural network for detection of coronavirus (SARS-CoV-2) cases from CT and Chest X-Ray images

“…The mathematical output of the convolution layer is given in formula 8 . The use of multiple convolution layers in deep convolutional neural network architectures causes an increase in the number of parameters, processing power, and cost [85] . In case of an increase in the number of convolution layers during the model's development process, the input data should also increase for the model to be trained efficiently.…”

“…Here, depthwise separable convolution, a promising technology, can be used as it uses fewer parameters than the standard convolution layer and reduces the processing volume in the calculation part. Depthwise separable convolution has been implemented in the MobileNet architecture and has produced successful results [85] . Depthwise convolution, separable convolution, and standard convolution study methodology are given in Fig.…”

“…Standard convolution processing cost is calculated as W K ∗ H K ∗ C in ∗ C out [85] . As a result of this situation, the depthwise separable convolution standard convolution cost ratio is calculated as follows [85] .…”

“…Standard convolution processing cost is calculated as W K ∗ H K ∗ C in ∗ C out [85] . As a result of this situation, the depthwise separable convolution standard convolution cost ratio is calculated as follows [85] . where; W, H, C in , C out , C inter are width, height, number of input channels, number of output channels and intermediate channels respectively.…”

Food Odor Recognition via Multi-step Classification

“…However, these methods usually cannot achieve good performance because of the high non-linearity behind the model. Non-linear methods such as artificial neural networks (ANN) [5] and convolutional neural networks (CNN) [6] have been well-developed in recent years. Nonetheless, the applications or demonstrations of odor classification via neural network were not sufficient enough to make a holistic evaluation of the performance.…”