Electronic nose systems imitate the biological odor mechanism via a model nose with chemical sensors. The book presents four main issues related to electronic noses by introducing the basic structure of a system. It includes much related work and presents a set of solutions for the algorithmic prediction challenge, drift compensation, disturbance elimination, and discreteness correction problems.

The book includes five parts. Part 1 introduces the related concepts and algorithmic challenges. Part 2 explains different approaches to the prediction challenge, including heuristic neural nets, chaos-based neural nets, multilayer perceptron (MLP)-based concentration estimation, support vector machine (SVM)-based odor classification, odor recognition based on local kernel discriminant analysis, and an ensemble of classifiers. Part 3 presents learning models for drift compensation. Part 4 identifies methods for dealing with disturbance elimination. Part 5 presents models for the discreteness challenge.

Chapter 1 explains the main electronic nose concepts. Chapter 2 then presents the challenges of electronic nose algorithms, including prediction, drift, disturbance, and discreteness.

Chapter 3--the first chapter on approaches to the prediction challenge--presents different particle swarm optimization techniques, two hybrid evolutionary algorithms, and how to apply those techniques on prediction mechanisms for concentration estimation by an electronic nose. It evaluates the effectiveness of the methods by reporting the estimation rates of chemical gases. While chapter 4 presents a chaotic optimization method on back-propagation neural networks for concentration estimation, chapter 5 introduces two multilayer perceptron-based quantization models for the same purpose by reporting both prediction errors and execution times of the algorithms. Chapter 6 evaluates the effectiveness of a hybrid SVM model for classifying multiple air contaminants. Chapter 7 proposes a local kernel discriminant analysis framework for multi-class recognition in an electronic nose system. Chapter 8 also aims to solve the multi-class recognition problem by proposing an SVM ensemble. While each chapter in Part 2 proposes different approaches for similar problems, the book lacks an overall evaluation. I was expecting to see all the accuracy and efficiency results together, in order to compare the proposed solutions.

Chapter 9 proposes a chaotic time-series prediction for sensor drift in electronic nose systems. Chapter 10 presents a drift compensation method based on an extreme learning machine by introducing domain adaptation. Chapter 11 introduces a drift removing/reducing method by applying a domain regularized dimension reduction algorithm. Chapter 12 proposes a cross-domain extreme learning machine model for drifted electronic nose data. Chapter 13 proposes another extreme learning machine model based on domain correction and transfer learning. Chapter 14 presents a multi-feature kernel semi-supervised joint learning model for complex and long-term drift.

Chapter 15 presents a least-squares SVM (LS-SVM) and an artificial neural network (ANN) for the discrimination of unwanted odor interference. Chapter 16 extends the work in chapter 5 by considering interference elimination as well as discrimination. It proposes pattern mismatch-based interference elimination for removing irrelevant gases from electronic nose sensor data. Chapter 17 proposes an extreme learning machine-based model for abnormal odor removal by defining the problem as a two-class classification.

Since electronic nose sensors may generate different responses for the same environment due to their insensitive profile over time or physical environment conditions, an electronic nose system needs to deal with this discreteness problem. Chapter 18 presents a calibration model for sensor responses, and ANN-based prediction methods for gas concentration. Chapter 19 extends the model given in chapter 18 for large-scale calibration by conducting experiments on different sensors. Chapter 20 summarizes the book and discusses future related work.

The book explains different strategies for dealing with the algorithmic challenges of electronic nose systems. Similar content is repeated in each chapter, for example, introducing basic concepts and related work, explaining the same base method, and so on. The chapters are not correlated--the electronic nose, data acquisition method, and dataset are different for each distinct system. It is good to see many studies in the area in a single resource; however, the content is available elsewhere.