Curriculum Learning-Based Approaches for End-to-End Gas Recognition

Zhang, Chao; Wang, Wen; Pan, Yong; Zhai, Shoupei

doi:10.1109/tim.2022.3205901

Cited by 5 publications

(1 citation statement)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Traditional methods of E-nose data processing are cumbersome and complex and require high-quality manual feature extraction, a process that makes recognition slow and cumbersome. To improve classification accuracy, artificial olfactory systems have improved in terms of sensors and classifiers, and researchers have proposed more artificial neural network algorithms [6], [7]. Recently, some researchers have also made use of deep learning algorithms for E-nose odor recognition.…”

Section: Introductionmentioning

confidence: 99%

A convolutional spiking neural network combined with residual blocks for electronic nose data processing

Zhang

et al. 2023

International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2023)

View full text Add to dashboard Cite

The traditional method of electronic nose (E-nose) data processing has the disadvantages of cumbersome operation steps and low classification accuracy. To address these problems, this paper proposes a convolutional spiking neural network (CSNN) for E-nose data processing that combines residual blocks. The network model consists of spiking-convolutional layers and fully connected pulse layers. The model combines the feature extraction capability of a convolutional neural network (CNN) with the computational efficiency of a spiking neural network (SNN) and the good biointerpretability of spike signal data and makes use of residual blocks to allow the network to learn richer content. In addition, two spike coding methods (response rate coding and response value coding) are designed to encode the data to make great use of the sensor curve features. To test the performance of the proposed network model in the E-nose, the data collected by the self-built E-nose system were used to identify and classify ten toxic gases with a maximum classification rate of 96.39%.

show abstract

Section: Introductionmentioning

confidence: 99%

A convolutional spiking neural network combined with residual blocks for electronic nose data processing

Zhang

et al. 2023

International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2023)

View full text Add to dashboard Cite

show abstract

Advances in the surface acoustic wave sensors for industrial applications: Potentials, challenges, and future directions: A review

Aslam,

Zhang,

Sreejith

et al. 2023

Measurement

View full text Add to dashboard Cite

MGRBA: Gas Recognition With Biclustering and Adaboost

Zhou,

Wang,

Huang

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Gas recognition has been widely used in many fields such as air quality monitoring in dangerous areas. However, existing recognition methods suffer from two limitations: first, the recognition accuracy is not high. Due to the stochastic nature of air turbulence, gas features are not steady. The global features are sensitive to feature variations. Existing methods are based on global similarity, ignoring local similarity. Samples may be dissimilar in respect of global similarity, but are similar in terms of local similarity; Second, most existing recognition methods are based on the closed-set assumption that the gases categories in the train and test set are same. However, in real world applications, the test set may have non-overlapping gas category with the train set. To address above limitations, biclustering is used to extract local similarity. However, original biclustering method is not suitable for extraction. Since original biclustering method is used to find all kinds of biclusters, here we just want to find column nearly constant bicluster Therefore, a modified biclustering method is proposed. The local similarity can be used to construct classifier to recognize gas with adaboost. However, original adaboost cannot be used for open-set recognition. Thus a modified adaboost that uses two thresholds is proposed to recognize the unknown gases. To assess the efficacy of the proposed method, it is tested on public dataset. Experiment results demonstrate that the proposed method outperforms several state-of-the-art methods in respect of several evaluation measures on both closed-set and open-set cases.

show abstract

Curriculum Learning-Based Approaches for End-to-End Gas Recognition

Cited by 5 publications

References 30 publications

A convolutional spiking neural network combined with residual blocks for electronic nose data processing

A convolutional spiking neural network combined with residual blocks for electronic nose data processing

Advances in the surface acoustic wave sensors for industrial applications: Potentials, challenges, and future directions: A review

MGRBA: Gas Recognition With Biclustering and Adaboost

Contact Info

Product

Resources

About