A neural approach for improving the measurement capability of an electronic nose

Recently, a growing interest in artificial implementations of biological systems has been arising. In particular, several research groups have been working in mimicking the mammalian olfactory system with the so-called electronic noses (e-noses). The e-noses, which are based on a sensor array, a fluid-dynamic system, and a data processing unit, are systems devoted to detecting and analyzing volatiles, where a deep knowledge of the target application is needed. In order to achieve effective results the sampling system, the measurement protocols, the sensor array, and the pattern recognition techniques have to be carefully designed. The increasing complexity of such design poses issues in sensory feature extraction and fusion, drift compensation, and data processing, especially when high efficiency is required for real-time applications. The interconnection and cooperation of several modules devoted to processing different tasks, such as control, data acquisition, data filtering interfaces, feature selection, and pattern analysis, are already mandatory. Moreover, heterogeneous techniques used to implement such tasks may introduce module interconnection and cooperation issues.In this paper, we address the development of a dedicated instrument able to perform real-time transduction, fusion, and processing of chemoresistive sensor array signals. In particular, this instrument realizes a dynamic and efficient management of data processing techniques and automatically controls the measurement protocols and the sampling system. An array of conducting poly (alkoxy-bithiophenes) sensors, the fluid-dynamic system, the electronic section, the framework's base architecture, and the implementation of dedicated application processes are described. The classification task is based on a self-organizing map where models for artificial neurons and connections were derived from the base structures available in the framework core. According to the target application, this instrument is portable and easily tailored, calibrated, and trained. Classification of olive oil headspaces supports its utility in supplying high-efficiency routine for volatile organic compounds detection and analysis.

show abstract

“…In order to discriminate among samples of the dataset, a neural network approach was adopted [24]. A 3 3 KSOM [25] was applied.…”

Section: Software Architecturementioning

confidence: 99%

“…Features constitute the data set for the section devoted to the analysis and classification of volatiles. A preliminary investigation with reference products responsible for olive oil aromas was previously reported [24].…”

mentioning

confidence: 99%

Towards a Real-Time Transduction and Classification of Chemoresistive Sensor Array Signals

2007

View full text Add to dashboard Cite

show abstract

“…Occupation of binding sites may come from water molecules of ambient air interacting with dopants or where headspace molecules were not desorbed properly. Purging with an inert gas such as nitrogen has been used by some authors to improve desorption and repeatability [7,31]. The reasons why responses in this study were in the form of a sine wave, however, remain unclear.…”

Section: Conducting Polymer E-nose Characterisationmentioning

confidence: 93%

“…Impedance was decreased with an increase in humidity and temperature [10][11][12][13]. The temperature optimum was found to be 25 • C [7,24].…”

Section: The Effect Of Temperaturementioning

confidence: 94%

“…As for human noses, sensors show an overlapping specificity and therefore provide a non-quantitative sensor response, a so-called fingerprint, rather than a specific qualitative or quantitative answer. The lack of specificity is due to the fact that sensors react with functional groups and structures of analytes rather than specifically with the molecule itself [7,8]. Sensor responses are then analysed using multivariate data analysis techniques in order to classify or discriminate a group of samples.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Methodological variation in headspace analysis of liquid samples using electronic nose

Knobloch

Turner

Spooner

et al. 2009

Sensors and Actuators B: Chemical

View full text Add to dashboard Cite

a b s t r a c tIn past years, numerous electronic nose (e-nose) developments have been published describing analyses of solid-, liquid-or gaseous media in microbiological-, environmental-, agricultural-or medical applications. However, little has been reported about complex methodological pitfalls that might be associated with commercially available e-nose technology. In this paper, some of these pitfalls such as temperature, the use of filters and mass flow using different sampling methods (static-and dynamic sampling) are described for two generations of conducting polymer e-noses (ST114/214, CPs, both Scensive Tech. Ltd.). A comparison with metal oxide semiconducting field effect transistor/metal oxide semiconductor (MOSFET/MOS) e-noses regarding stability across replicates and over time was made. Changes in temperature were found to give larger sensor responses, whereas the application of filters led to quantitative and qualitative changes in sensor responses due to a change in mass flow which was also affected by the sampling method. Static sampling provided more stable flows across replicates. Variation was investigated for CPs and MOSFET/MOS e-noses that gave different responses over time and across replicates. These methodological factors cause a lack of stability and reproducibility, demonstrating the pitfalls of e-nose technology and therefore limit their utility for discriminating between samples.

show abstract

Application of sensor dynamic response analysis to improve the accuracy of odour-measuring systems

Sobański¹,

Modrak²,

Nitsch³

et al. 2005

Meas. Sci. Technol.

View full text Add to dashboard Cite

A system consisting of a matrix of three semiconductor gas sensors was applied to the classification of different orange juices. The sensor matrix responses were sampled in short time intervals. Such responses were processed by discrete wavelet transform (DWT) together with the k-nearest neighbour (kNN) classification algorithm or by the probabilistic neural network (PNN). The obtained results show that both types of signal processing (DWT + kNN and PNN) applied provide very good class separation for time response analysis, while in the case of the static response analysis the correct classification coefficients are much lower. It is shown that the analysis of the sensor's time response can be an efficient way of increasing both the accuracy level and the immunity to external noise in e-nose systems. The possibility of reducing the number of sensors without decreasing the system performance is also demonstrated. Additional experiments have shown that for both processing methods, the results obtained with the dynamic response of a single sensor were better than those reached with the three-sensor array measured in static conditions.

show abstract

A neural approach for improving the measurement capability of an electronic nose

Cited by 8 publications

References 12 publications

Towards a Real-Time Transduction and Classification of Chemoresistive Sensor Array Signals

Towards a Real-Time Transduction and Classification of Chemoresistive Sensor Array Signals

Methodological variation in headspace analysis of liquid samples using electronic nose

Application of sensor dynamic response analysis to improve the accuracy of odour-measuring systems

Contact Info

Product

Resources

About