Electronic noses (e-nose) are composed of an appropriate pattern recognition system and a gas sensor array with a certain degree of specificity and broad spectrum characteristics. The gas sensors have their own shortcomings of being highly sensitive to interferences which has an impact on the detection of target gases. When there are interferences, the performance of the e-nose will deteriorate. Therefore, it is urgent to study interference suppression techniques for e-noses. This paper summarizes the sources of interferences and reviews the advances made in recent years in interference suppression for e-noses. According to the factors which cause interference, interferences can be classified into two types: interference caused by changes of operating conditions and interference caused by hardware failures. The existing suppression methods were summarized and analyzed from these two aspects. Since the interferences of e-noses are uncertain and unstable, it can be found that some nonlinear methods have good effects for interference suppression, such as methods based on transfer learning, adaptive methods, etc.

The long-term sensor drift phenomenon seriously restricts the performance of Electronic Nose (E-nose) systems in their various applications. Due to frequent recalibrations, traditional drift compensation methods are costly and laborious, and their performance are limited due to the nonlinear dynamic properties of the drift. The latest proposed Broad Learning System (BLS) has been confirmed to be an efficient and effective learning technique for many machine learning problems. However, BLS with cross-domain learning capability has rarely been studied. In this paper, a novel unified framework called Domain Transfer Broad Learning System (DTBLS) is proposed based on BLS, to address the issue of drift via adaptive compensation. For the case where there is no labeled target sample, with simultaneous considerations of the empirical loss of source data, marginal distribution adaptation, conditional distribution adaptation and manifold regularization, the DTBLS framework learns a robust target classifier by using labeled source data and unlabeled target data to compensate the drift of sensor response adaptively. To the best of our knowledge, DTBLS is the first BLS-based transfer learning framework for the problem of dataset shift existing in E-nose systems. Like the basic BLS, high computation efficiency is achieved due to the existence of analytical solution. Parameter sensitivity analysis is also conducted to show that the optimal solution can be obtained in a wide range. Experiments on a public gas sensor drift dataset demonstrate that the proposed method outperforms the state-of-the-art methods well.

Near-infrared (NIR) spectral sensors can deliver the spectral response of light absorbed by materials. Data analysis technology based on NIR sensors has been a useful tool for quality identification. In this paper, an improved deep convolutional neural network (CNN) with batch normalization and MSRA (Microsoft Research Asia) initialization is proposed to discriminate the tobacco cultivation regions using data collected from NIR sensors. The network structure is created with six convolutional layers and three full connection layers, and the learning rate is controlled by exponential attenuation method. One-dimensional kernel is applied as the convolution kernel to extract features. Meanwhile, the methods of L2 regularization and dropout are used to avoid the overfitting problem, which improve the generalization ability of the network. Experimental results show that the proposed deep network structure can effectively extract the complex characteristics inside the spectrum, which proves that it has excellent recognition performance on tobacco cultivation region discrimination, and it also demonstrates that the deep CNN is more suitable for information mining and analysis of big data.

In this article, we investigate a novel multisensor odor detection system (electronic nose) for low-concentration volatile organic compounds (VOCs). In order to break through the limitation for detecting low concentration VOCs, we design a conformal symmetric preconcentration unit structure based on the finite element method. A typical application for exhaled breath is developed to investigate the proposed scheme. The designed preconcentration system can increase the concentration of the concentrated substance, so that the sensor can detect it. Besides, we choose the alveolar gas at the bottom of the human lung, which can better represent the health status of the human body, to carry out the effectiveness test of the system. In order to combine the gas acquisition system with the preconcentration system, this article further designs a cost-effective alveolar gas collection device by a doubled adaptive pulsewidth modulation (PWM) operation. Finally, the effectiveness of the preconcentration system is judged by the detection results of gas chromatography-mass spectrometry (GC-MS). We also use the classical algorithm to classify lung cancer and nonlung cancer patients, which proves the effectiveness of our proposed integrated design system and shows the potential of practical application of the system.

Near-infrared (NIR) spectral sensors deliver the spectral response of the light absorbed by materials for quantification, qualification or identification. Spectral analysis technology based on the NIR sensor has been a useful tool for complex information processing and high precision identification in the tobacco industry. In this paper, a novel method based on the support vector machine (SVM) is proposed to discriminate the tobacco cultivation region using the near-infrared (NIR) sensors, where the genetic algorithm (GA) is employed for input subset selection to identify the effective principal components (PCs) for the SVM model. With the same number of PCs as the inputs to the SVM model, a number of comparative experiments were conducted between the effective PCs selected by GA and the PCs orderly starting from the first one. The model performance was evaluated in terms of prediction accuracy and four parameters of assessment criteria (true positive rate, true negative rate, positive predictive value and F1 score). From the results, it is interesting to find that some PCs with less information may contribute more to the cultivation regions and are considered as more effective PCs, and the SVM model with the effective PCs selected by GA has a superior discrimination capacity. The proposed GA-SVM model can effectively learn the relationship between tobacco cultivation regions and tobacco NIR sensor data.