Minghua Shi

Fast and robust gas identification system using an integrated gas sensor technology and Gaussian mixture models

Among the most serious limitations facing the success of future consumer gas identification systems are the drift problem and the real-time detection due to the slow response of most of todays gas sensors. This paper shows that the combination of an integrated sensor array and a Gaussian mixture model permits success in gas identification problems. An integrated sensor array has been designed with the aim of combustion gases identification. Our identification system is able to quickly recognize gases with more than 96% accuracy. Robust detection is introduced through a drift counteraction approach based on extending the training data set using a simulated drift.

An Efficient Digital VLSI Implementation of Gaussian Mixture Models-Based Classifier

Gaussian mixture models (GMM)-based classifiers have shown increased attention in many pattern recognition applications. Improved performances have been demonstrated in many applications, but using such classifiers can require large storage and complex processing units due to exponential calculations and a large number of coefficients involved. This poses a serious problem for portable real-time pattern recognition applications. In this paper, first the performance of GMM and its hardware complexity are analyzed and compared with a number of benchmark algorithms. Next, an efficient digital hardware implementation is proposed. A number of design strategies are proposed in order to achieve the best possible tradeoffs between circuit complexity and real-time processing. First, a serial-parallel vector-matrix multiplier combined with an efficient pipelining technique is used. A novel exponential calculation circuit based on a linear piecewise approximation is proposed to reduce hardware complexity. The precision requirement of the GMM parameters in our classifier are also studied for various classification problems. The proposed hardware implementation features programmability and flexibility offering the possibility to use the proposed architecture for different applications with different topologies and precision requirements. To validate the proposed approach, a prototype was implemented in 0.25-mum CMOS technology and its operation was successfully tested for gas identification application

An Efficient FPGA Implementation of Gaussian Mixture Models-Based Classifier Using Distributed Arithmetic

Gaussian mixture models (GMM)-based classifiers have shown increased attention in many pattern recognition applications. Improved performances have been demonstrated in many applications but using such classifiers can require large storage and complex processing units due to exponential calculations and large number of coefficients involved. This poses a serious problem for portable real-time pattern recognition applications. In this paper, first the performance of GMM and its hardware complexity are analyzed and compared with a number of benchmark algorithms. Next, an efficient digital hardware implementation based on distributed arithmetic (DA) is proposed. A novel exponential calculation circuit based on linear piecewise approximation is also developed to reduce hardware complexity. Implementation is carried out on the Celoxica-RC1000 board equipped with the Virtex-E FPGA. Maximum optimization has been achieved by means of manual placement and routing in order to achieve a compact core footprint. A detailed evaluation of the performance metrics of the GMM core is also presented.

Performance Enhanced Voltage Scaling in FPGAs

As field programmable gate array (FPGA) based systems scale up in complexity, energy aware designs paradigms with strict power budgets require the designer to explore all viable options for minimising dynamic power consumption. The concepts of parallelism and pipelining have long been exploited in CMOS chips to reduce power and energy consumption. In this paper, a systematic empirical study of the tradeoffs between degree of parallelism, threshold voltage and power consumption under constant throughput conditions commercially available FPGAs has been presented. Results indicate that there is excellent scope for reduction in dynamic voltage by suitably applying the tradeoffs in FPGA based designs in order to achieve energy efficient implementations.

Redundancy Analysis for Tin Oxide Gas Sensor Array

Using gas sensor array is widely accepted to overcome the non-selectivity of a single sensor. For tin oxide gas sensors, the size of array can’t be very large due to the limited number of doping materials. In this paper, our experimental results shows that duplication of the sensors doped by the same metal is an efficient way to improve the selectivity of the array due to the fabrication mismatch of the sensor chip. We also compare two methods of reducing the dimension of gas patterns: removing the sensors providing redundant information in the array and using principle component analysis (PCA). The experimental results shows that when the number of components is too large PCA can be a useful tool to reduce the data dimension.

Quantization errors in committee machine for gas sensor applications

In a digital implementation of a gas identification system, the mapping of continuous real parameter values into a finite set of discrete values introduces an error into the system. This paper presents the results of an investigation into the effects of parameter quantization on different classifiers (KNN, MLP and GMM). We propose a committee machine to decrease the classification performance degradation due to the quantization errors. The simulation results show that the committee machine always outperforms a single classifier and the gain in classification performance is greater for a reduced number of bits.

FPGA Based Run Time Reconfigurable Gas Discrimination System

In this paper a gas discrimination system based on five classification algorithms including K nearest neighbors (KNN), multi-layer perceptron (MLP), radial basis function (RBF), Gaussian mixture model (GMM) and probabilistic principal component analysis (PPCA) has been presented. A Committee machine (CM) is used in which the results from each classifier are first transformed to confidences and then a weighted combination rule is used to generate the final decision result. In order to overcome the problem of very high computational complexity of the CM requiring large amount of hardware resources, we propose a novel time multiplexing hardware implementation using a dynamically reconfigurable field programmable gate array (FPGA) platform. The system is successfully tested for combustible gas identification application using our in-house tin-oxide gas sensors.