E-nose System Research Articles

Enhanced Diabetes Detection and Blood Glucose Prediction Using TinyML-Integrated E-Nose and Breath Analysis: A Novel Approach Combining Synthetic and Real-World Data

Diabetes mellitus, a chronic condition affecting millions worldwide, necessitates continuous monitoring of blood glucose level (BGL). The increasing prevalence of diabetes has driven the development of non-invasive methods, such as electronic noses (e-noses), for analyzing exhaled breath and detecting biomarkers in volatile organic compounds (VOCs). Effective machine learning models require extensive patient data to ensure accurate BGL predictions, but previous studies have been limited by small sample sizes. This study addresses this limitation by employing conditional generative adversarial networks (CTGAN) to generate synthetic data from real-world tests involving 29 healthy and 29 diabetic participants, resulting in over 14,000 new synthetic samples. These data were used to validate machine learning models for diabetes detection and BGL prediction, integrated into a Tiny Machine Learning (TinyML) e-nose system for real-time analysis. The proposed models achieved an 86% accuracy in BGL identification using LightGBM (Light Gradient Boosting Machine) and a 94.14% accuracy in diabetes detection using Random Forest. These results demonstrate the efficacy of enhancing machine learning models with both real and synthetic data, particularly in non-invasive systems integrating e-noses with TinyML. This study signifies a major advancement in non-invasive diabetes monitoring, underscoring the transformative potential of TinyML-powered e-nose systems in healthcare applications.

A Portable Electronic Nose Coupled with Deep Learning for Enhanced Detection and Differentiation of Local Thai Craft Spirits

In this study, our primary focus is the biomimetic design and rigorous evaluation of an economically viable and portable ‘e-nose’ system, tailored for the precise detection of a broad range of volatile organic compounds (VOCs) in local Thai craft spirits. This e-nose system is innovatively equipped with cost-efficient metal oxide gas sensors and a temperature/humidity sensor, ensuring comprehensive and accurate sensing. A custom-designed real-time data acquisition system is integrated, featuring gas flow control, humidity filters, dual sensing/reference chambers, an analog-to-digital converter, and seamless data integration with a laptop. Deep learning, utilizing a multilayer perceptron (MLP), is employed to achieve highly effective classification of local Thai craft spirits, demonstrated by a perfect classification accuracy of 100% in experimental studies. This work underscores the significant potential of biomimetic principles in advancing cost-effective, portable, and analytically precise e-nose systems, offering valuable insights into future applications of advanced gas sensor technology in food, biomedical, and environmental monitoring and safety.

Low-power adaptive sampling electronic nose system with a Radon transform-based convolutional neural network for optimized gas recognition

The electronic nose (E-nose) systems provide great promise in capturing information about the identity of gaseous chemical analytes as well as their temporal variation. However, the regular and dense sampling events of the E-nose often lead to a substantial amount of redundancy in the temporal structure, which in turn limits their application in efficiency-sensitive scenarios such as advanced robotics and the Internet of Things. Herein, we introduce a novel operating mechanism for E-noses, which dynamically adjusts the time interval of operation to reduce power consumption and information content through an adaptive sampling mechanism. Three strategic schemes with different operating principles are presented to implement this mechanism: the accurate scheme targets the optimized discrimination accuracy, the vague scheme is oriented toward power-sensitive scenarios, and the balanced scheme aims to balance cost and benefit. In order to ensure that the adaptively sampled inhomogeneous data can be compatible with each other, we further propose an algorithm based on the Radon transform that can convert non-uniform time series with timestamp information into the same size tensor. Finally, we designed a convolutional neural network-based classification model for gaseous chemical analytes, providing guidance on scheme and parameter selection with an accuracy of up to 99 % and power savings of up to 96 %. Overall, this work provides a novel solution to optimize temporal redundancy in E-nose systems and a generalized approach to the corresponding deep-learning data processing.

Enhancing mixed gas discrimination in e-nose system: sparse recurrent neural networks using transient current fluctuation of SMO array sensor

Despite recent significant advancements in gas sensor array technology, accurately identifying gases in mixed environments remains challenging. This difficulty is primarily due to the rapid and competing processes of gas molecules attaching to (adsorption) and detaching from (desorption) the sensor. In this study, we present a simple method to fabricate a 2 × 4 SMO-based gas sensor array, coupled with a sparse recurrent neural network (SRNN) that employs weight regularization. The recurrent layers of the SRNN process nonlinear information and capture temporal dependencies in the sensor data, while the regularization technique simplifies the model, making it both efficient and easier to interpret. Additionally, we introduce a novel feature: the dynamics of current, labeled as ΔI. This feature enables the SRNN model to efficiently detect the adsorption and desorption of gas molecules. We demonstrate that our model can distinguish between three intuitively indistinguishable datasets of gas species (NO2, HCHO, and a mixture) with up to 92 % accuracy. By utilizing the fast and competitive adsorption/desorption information of gas molecules, our model can be applied to various gas combination environments, unlike conventional gas sensing data measured over longer periods. By integrating the sensor array with the advanced SRNN model, we pave the way for sophisticated e-nose systems, with potential applications in advanced gas sensing technologies, such as disease diagnosis through exhaled breath analysis and the detection of toxic species in mixed gas environments.

Low-Cost Electronic Nose for Robusta and Arabica Coffee Classification Using 1-D Convolutional Neural Networks

Abstract The use of an electronic nose (e-nose) employing MQ-series sensors has become increasingly popular due to its cost-effectiveness. However, the impact of varying sampling durations on its performance, particularly in distinguishing between Robusta and Arabica coffee, has not been extensively studied. This research investigates how different sampling durations affect the e-nose’s classification accuracy. The study utilizes a 1D Convolutional Neural Network (1D-CNN) and a Support Vector Machine (SVM), with both models enhanced by a Savitzky-Golay filter to reduce noise and improve data quality. Feature selection techniques are applied to address data complexity and improve model performance. The experimental results indicate that the 1D-CNN model achieves optimal accuracy of up to 100% at a sampling duration of 200 seconds, while the SVM reaches approximately 92% accuracy under the same conditions. Notably, for applications requiring faster processing times, the SVM proves to be more effective, achieving 70% accuracy at a 20-second sampling interval, compared to 60% for the 1D-CNN. The study demonstrates that integrating MQ sensors in e-nose systems can yield effective classification results for coffee varieties, even with shorter sampling times. These findings have the potential to influence the development of cost-effective e-nose devices, making them more accessible to small and medium-sized industrial enterprises.

Ripeness Assessment and Quality Control of Mango Gold Susu using an E-Nose System

Ripeness Assessment and Quality Control of Mango Gold Susu using an E-Nose System

An electronic nose drift compensation algorithm based on semi-supervised adversarial domain adaptive convolutional neural network

Sensor drift is a significant challenge leading to performance degradation in electronic nose(E-nose) systems. Effectively addressing sensor drift represents the most daunting problem in E-nose technology. This work proposes a domain adaptive approach called Semi-Supervised Adversarial Domain Adaptive Convolutional Neural Network (SAD-CNN) to tackle the long-term drift in E-nose and device displacement. SAD-CNN leverages adversarial learning to minimize distribution disparities between the source and target domains. Unlike traditional methods employing projection matrices, SAD-CNN utilizes one-dimensional convolutional neural networks as feature extractors, circumventing the complexities associated with parameter adjustments and matrix calculations in the projection process. During the training process, utilizing pseudo-labels generated by the semi-supervised self-training method to train the model, thereby reducing the labeling costs. Additionally, confidence threshold screening is introduced during the self-training phase to minimize erroneous pseudo-labels. Furthermore, the regenerative Hilbert space’s Maximum Mean Difference combined with Minimum Variance is introduced as a domain constraint function to mitigate distribution discrepancies between domains and enhance feature discriminability across domains. The experimental results demonstrate that the SAD-CNN method outperforms others. Within the long-term drift dataset, the classification accuracies for different scenarios are 78.01 % and 82.53 %, respectively. Meanwhile, the instrument change dataset yields a classification accuracy of 96.45 %.

Sensory Perception Systems and Machine Learning Methods for Pesticide Detection in Fruits

In this study, an electronic tongue (E-tongue) and electronic nose (E-nose) systems were applied to detect pesticide residues, specifically Preza, Daconil, Curzate, Bricol, Accros, Amistar, and Funlate, in fruits such as cape gooseberries, apples, plums, and strawberries. These advanced systems present several advantages over conventional methods (e.g., GC-MS and others), including faster analysis, lower costs, ease of use, and portability. Additionally, they enable non-destructive testing and real-time monitoring, making them ideal for routine screenings and on-site analyses where effective detection is crucial. The collected data underwent rigorous analysis through multivariate techniques, specifically principal component analysis (PCA) and linear discriminant analysis (LDA). The application of machine learning (ML) algorithms resulted in a good outcome, achieving high accuracies in identifying fruits contaminated with pesticides and accurately determining the concentrations of those pesticides. This level of precision underscores the robustness and reliability of the methodologies employed, highlighting their potential as alternative tools for pesticide residue detection in agricultural products.

Discrimination of VOCs with e-nose consisting of a single MEMS sensor using lightweight deep learning model based on SqueezeNet

The detection of volatile organic compounds (VOCs) has numerous applications in environmental monitoring and biomedicine. Metal oxide semiconductor (MOS) gas sensors are widely used in gas detection due to their simple structure, low fabrication cost, and fast response speed. However, MOS gas sensors have broad-spectrum responsiveness to VOCs, making selectivity a challenge. The most commonly used method to obtain selectivity is the E-nose system consisting of multiple sensors. However, the size of E-noses has increased due to the number of sensors with high power consumption and complicated signal processing. In recent years, deep learning has flourished, allowing for accurate discrimination of VOCs and concentration prediction using E-nose consisting of a single sensor. This method significantly reduces power consumption and facilitates portable applications. In this work, we synthesize SnO2-ZnO composite oxide and fabricate a E-nose consisting of a single MEMS gas sensor. The response performance of the E-nose to VOCs is stable with a high output signal-noise ratio (SNR). We design a lightweight deep learning model based on SqueezeNet and train it using transfer learning. Linear regression is used to predict the concentration of VOCs. Our work achieves accurate identification and low-error concentration prediction of VOCs, which is valuable for subsequent portable VOC detection systems.

Advances in drift compensation algorithms for electronic nose technology

Purpose This paper aims to explore recent advances in drift compensation algorithms for Electronic Nose (E-nose) technology and addresses sensor drift challenges through offline, online and neural network-based strategies. It offers a comprehensive review and covers causes of drift, compensation methods and future directions. This synthesis provides insights for enhancing the reliability and effectiveness of E-nose systems in drift issues. Design/methodology/approach The article adopts a comprehensive approach and systematically explores the causes of sensor drift in E-nose systems and proposes various compensation strategies. It covers both offline and online compensation methods, as well as neural network-based approaches, and provides a holistic view of the available techniques. Findings The article provides a comprehensive overview of drift compensation algorithms for E-nose technology and consolidates recent research insights. It addresses challenges like sensor calibration and algorithm complexity, while discussing future directions. Readers gain an understanding of the current state-of-the-art and emerging trends in electronic olfaction. Originality/value This article presents a comprehensive review of the latest advancements in drift compensation algorithms for electronic nose technology and covers the causes of drift, offline drift compensation algorithms, online drift compensation algorithms and neural network drift compensation algorithms. The article also summarizes and discusses the current challenges and future directions of drift compensation algorithms in electronic nose systems.

Prediction of Mango Quality During Ripening Stage Using MQ-based Electronic Nose and Multiple Linear Regression

In recent years, consumers have shown interest in non-destructive methods to assess the fruit's internal quality during ripening. The objective of this study is to construct and evaluate an E-Nose system using low-cost MQ sensors to evaluate fruit quality, specifically soluble sugar content (SSC) and hardness of mango during ripening. The correlation test was performed to compare sensor readings with SSC and hardness, and multiple linear regression (MLR) was used to establish linear equations for mango quality indices based on sensor variation. Over the storage period, the hardness of mango was decreased from the value of 15.4 kgf/cm² to 12.25 kgf/cm². Similarly, the SSC for mangoes increased from 19.7%Brix to a final value of 24.66%Brix. The sensor values also showed positive correlation with SSC and negative correlation with hardness of mango, respectively. Using the MLR analysis, the hardness and SSC of mango during the ripening stage, the correlation coefficient (r) of 0.847, standard error of 1.49 kgf/cm2 and 0.815, standard error of 1.696 % Brix for hardness and SSC prediction, respectively. These results indicate that MQ-based E-nose is the rapid and non-destructive method for predicting mango qualities during ripening stages.

Flavor identification based on olfactory-taste synesthesia model and hybrid convolutional neural network-random forest

Abstract The bionic-based electronic nose (e-nose) and electronic tongue (e-tongue) show satisfactory performance in flavor analysis. Traditional flavor analysis of the e-nose and e-tongue systems focuses on data fusion, and the effects of the bionic characteristics on the flavor analysis performance are rarely studied. Motivated by this, a method, including an olfactory-taste synesthesia model (OTSM) and a convolutional neural network-random forest (CNN-RF), is proposed for the effective identification of flavor substances. The OTSM is developed for human nerve conduction mechanisms to enhance the bionic characteristics of the e-nose and e-tongue systems and is combined with a CNN-RF model for flavor identification. The results show that, first, when stimulated by e-nose and e-tongue data, physiological 1/f characteristics and synchronization are shown using the OTSM. The enhancement effects on the bionic characteristics of the fusion system are validated using the 1/f characteristics and synchronization. Second, the fully connected layer for the CNN is replaced by RF to improve the identification performance of flavor substances. Finally, CNN-RF is evaluated in comparison with other flavor recognition models and ablation studies to confirm its effectiveness. By comparison, the best recognition performance, including the accuracies of 96.67%, 96.67%, and 95.00%, the F1-scores of 96.65%, 96.66%, and 94.95%, and the kappa coefficients of 96.03%, 96.10%, and 93.44%, for five beers, five apples, and four mixed solutions, respectively, is obtained by CNN-RF. In conclusion, excellent flavor identification for the fusion system is achieved using the OTSM and CNN-RF models.

Quality characterization of tobacco flavor and tobacco leaf position identification based on homemade electronic nose

A set of nine unique tobacco extract samples was analyzed using a self-developed electronic nose (E-nose) system, a commercial E-nose, and gas chromatography-mass spectrometry (GC–MS). The evaluation employed principal component analysis, statistical quality control, and soft independent modeling of class analogies (SIMCA). These multifaceted statistical methods scrutinized the collected data. Subsequently, a quality control model was devised to assess the stability of the sample quality. The results showed that the custom E-nose system could successfully distinguish between tobacco extracts with similar odors. After further training and the development of a quality control model for accepted tobacco extracts, it was possible to identify samples with normal and abnormal quality. To further validate our E-nose and extend its use within the tobacco industry, we collected and accurately classified the flavors of different tobacco leaf positions, with a remarkable accuracy rate of 0.9744. This finding facilitates the practical application of our E-nose system for the efficient identification of tobacco leaf positions.

Development of a Neural Network for Target Gas Detection in Interdigitated Electrode Sensor-Based E-Nose Systems.

In this study, a neural network was developed for the detection of acetone, ethanol, chloroform, and air pollutant NO2 gases using an Interdigitated Electrode (IDE) sensor-based e-nose system. A bioimpedance spectroscopy (BIS)-based interface circuit was used to measure sensor responses in the e-nose system. The sensor was fed with a sinusoidal voltage at 10 MHz frequency and 0.707 V amplitude. Sensor responses were sampled at 100 Hz frequency and converted to digital data with 16-bit resolution. The highest change in impedance magnitude obtained in the e-nose system against chloroform gas was recorded as 24.86 Ω over a concentration range of 0-11,720 ppm. The highest gas detection sensitivity of the e-nose system was calculated as 0.7825 Ω/ppm against 6.7 ppm NO2 gas. Before training with the neural network, data were filtered from noise using Kalman filtering. Principal Component Analysis (PCA) was applied to the improved signal data for dimensionality reduction, separating them from noise and outliers with low variance and non-informative characteristics. The neural network model created is multi-layered and employs the backpropagation algorithm. The Xavier initialization method was used for determining the initial weights of neurons. The neural network successfully classified NO2 (6.7 ppm), acetone (1820 ppm), ethanol (1820 ppm), and chloroform (1465 ppm) gases with a test accuracy of 87.16%. The neural network achieved this test accuracy in a training time of 239.54 milliseconds. As sensor sensitivity increases, the detection capability of the neural network also improves.

An E-nose system for identification and quantification of hazardous gas mixtures using a combined strategy of CNNs and attentional mechanisms

The chemical industry generates a broad spectrum of hazardous gases, presenting significant challenges for conventional detection methods due to their diverse chemical properties and low concentration levels. E-nose systems, employing sensor arrays, offer significant potential for the determination of gas mixtures. This study presents a novel E-nose algorithm, CNN-ECA, which integrated CNNs and attention mechanisms to improve the recognition accuracy of E-nose systems. By integrating the attention mechanism module into CNN’s convolutional operations, the algorithm emphasizes critical feature information. Three hazardous gases (ammonia, methanol, and acetone) and their mixtures were chosen as target gases. CNNs were combined with various attention mechanism networks (SENet, ECA, and CBAM) to construct models, which were then employed to train and evaluate data collected from the sensor array. The results were compared with traditional network models (KNN, SVM, and CNN). Experimental findings indicated that the prediction performance of CNN models combined with attention mechanism networks surpassed that of traditional network models. Particularly, the CNN-ECA network model demonstrated the highest performance in both qualitative and quantitative analyses. This study presents a promising solution for mixed gas detection by synergizing CNN and attention mechanism networks, thereby enhancing the accuracy and reliability of mixed gas measurements. Moreover, capitalizing on the lightweight architecture of the CNN-ECA model, transfer learning techniques were employed to adapt it for deployment on the Raspberry Pi hardware platform. This facilitates the development of a real-time E-nose system for gas detection.

Identification of the quality for rice with different storage humidity: An electronic nose combined with multiblock feature integration method

The ability of traditional attention mechanisms (AMs) to extract useful features from electronic nose (e-nose) data is limited, which affects the performance of the e-nose system for the identification of rice quality. Motivated by this, a nondestructive testing method incorporating an e-nose and multiblock feature integration (MBFI) is proposed to effectively discriminate the quality of rice at different storage humidity. First, gas information for two brands of rice at five storage humidity is acquired using the e-nose system. Second, the feature-mining ability of quality classification models is enhanced by the MBFI module. Finally, compared with the recognition results of multiple AMs, multiple classification models, and ablation analysis, the best identification performance and stability for rice quality are obtained by the MBFI and a residual network18 model. In conclusion, effective identification of rice quality is achieved by the e-nose and MBFI.

E-Nose: Time-Frequency Attention Convolutional Neural Network for Gas Classification and Concentration Prediction.

In the electronic nose (E-nose) systems, gas type recognition and accurate concentration prediction are some of the most challenging issues. This study introduced an innovative pattern recognition method of time-frequency attention convolutional neural network (TFA-CNN). A time-frequency attention block was designed in the network, aiming to excavate and effectively integrate the temporal and frequency domain information in the E-nose signals to enhance the performance of gas classification and concentration prediction tasks. Additionally, a novel data augmentation strategy was developed, manipulating the feature channels and time dimensions to reduce the interference of sensor drift and redundant information, thereby enhancing the model's robustness and adaptability. Utilizing two types of metal-oxide-semiconductor gas sensors, this research conducted qualitative and quantitative analysis on five target gases. The evaluation results showed that the classification accuracy could reach 100%, and the coefficient of the determination (R2) score of the regression task was up to 0.99. The Pearson correlation coefficient (r) was 0.99, and the mean absolute error (MAE) was 1.54 ppm. The experimental test results were almost consistent with the system predictions, and the MAE was 1.39 ppm. This study provides a method of network learning that combines time-frequency domain information, exhibiting high performance in gas classification and concentration prediction within the E-nose system.

Indirect evaluation of watermelon volatile profile: Detection of subtle changes with e-nose systems

The effectiveness of e-nose systems as high-throughput tools for volatile profiling in watermelon was investigated focusing on discerning subtle changes induced by the use of different rootstocks. Partial Least Square Discriminant Analysis (PLS-DA) models, both GC-MS and e-nose data, demonstrated moderate performance in classification due to nuanced differences among groups (the same F1 hybrid was used as scion). However, PLS-DA biplots revealed a clear correlation between GC-MS and e-nose data. This methodology enabled the e-nose system to identify the effects of specific root-scion combinations compared to non-grafted controls and detect combinations with more variable volatile profiles. Remarkably, the e-nose system identified samples with anomalous volatile profiles, mirroring the capabilities of GC-MS data. Additionally, PLS models were developed to provide reasonably accurate predictions of key compound contents like geranylacetone, (Z)-6-nonen-1-ol, or (Z)-6-nonenal, crucial for watermelon flavor and taste perception. Overall, this study highlights the potential of e-nose systems in discerning nuanced variations in watermelon volatile profiles affecting aroma. Incorporating volatile profile evaluation capabilities using such systems will significantly optimize quality control processes and plant breeding programs.

Exploring flavor perception through metabolite profiling and sensory approaches during starter kimchi fermentation

Starter cultures can enhance sensory properties during the starter kimchi fermentation. However, while many studies have primarily focused on physicochemical properties, the impact of starters on flavor perception remains unclear. The aim of this study was to assess the impact of starter culture inoculation, specifically Companilactobacillus allii WiKim39 and Lactococcus lactis WiKim0124, on aroma and flavor perception through a comparison of metabolite profiling and electronic tongue and nose data. Notably, 19 and 11 compounds were identified using UHPLC-Q-Orbitrap-MS and HPLC, respectively. Principal component analysis revealed that variations in fermentation metabolites and sensory properties were significantly categorized based on the particular starter culture employed during the ripening process. Moreover, the correlation analysis between the 19 key fermentative metabolites and e-tongue measurements in each starter kimchi revealed significant differences in sourness (AHS), bitterness (SCS), and comprehensive taste (CPS and PKS) sensors. The e-nose system identified four major volatile compounds contributing to the sweetness and off-flavor-related aroma profile of the starter kimchi. These findings highlight the significant impact that starter culture has on taste and flavor development, serving as a valuable guide in predicting sensory properties during starter kimchi fermentation.

Ultralow-Power Single-Sensor-Based E-Nose System Powered by Duty Cycling and Deep Learning for Real-Time Gas Identification.

This study presents a novel, ultralow-power single-sensor-based electronic nose (e-nose) system for real-time gas identification, distinguishing itself from conventional sensor-array-based e-nose systems, whose power consumption and cost increase with the number of sensors. Our system employs a single metal oxide semiconductor (MOS) sensor built on a suspended 1D nanoheater, driven by duty cycling─characterized by repeated pulsed power inputs. The sensor's ultrafast thermal response, enabled by its small size, effectively decouples the effects of temperature and surface charge exchange on the MOS nanomaterial's conductivity. This provides distinct sensing signals that alternate between responses coupled with and decoupled from the thermally enhanced conductivity, all within a single time domain during duty cycling. The magnitude and ratio of these dual responses vary depending on the gas type and concentration, facilitating the early stage gas identification of five gas types within 30 s via a convolutional neural network (classification accuracy = 93.9%, concentration regression error = 19.8%). Additionally, the duty-cycling mode significantly reduces power consumption by up to 90%, lowering it to 160 μW to heat the sensor to 250 °C. Manufactured using only wafer-level batch microfabrication processes, this innovative e-nose system promises the facile implementation of battery-driven, long-term, and cost-effective IoT monitoring systems.