Neural Network for Electronic Nose using Field Programmable Analog Arrays

Abstract

Electronic nose is a device detecting odors which is designed to resemble
the ability of the human nose, usually applied to the robot. The process of
identification of the electronic nose will run into a problem when the gas
which is detected has the same chemical element. Misidentification due to
the similarity of chemical properties of gases is possible; it can be solved
using neural network algorithms. The attendance of Field Programmable
Analog Array (FPAA) enables the design and implementation of an
analog neural network, while the advantage of analog neural network
which is an input signal from the sensor can be processed directly by the
FPAA without having to be converted into a digital signal. Direct analog
signal process can reduce errors due to conversion and speed up the
computing process. The small size and low power usage of FPAA are very
suitable when it is used for the implementation of the electronic nose that
will be applied to the robot. From this study, it was shown that the
implementation of analog neural network in FPAA can support the
performance of electronic nose in terms of flexibility (resource component
required), speed, and power consumption. To build an analog neural
network with three input nodes and two output nodes only need two
pieces of Configurable Analog Block (CAB), of the four provided by the
FPAA. Analog neural network construction has a speed of the process
0.375 μs, and requires only 59 ± 18mW resources.

DOI:http://dx.doi.org/10.11591/ijece.v2i6.1501