30 August 2021 in IEEE Transactions on Instrumentation and Measurement
The gas sensor drift problem arises from the bias of data, which is known as a significant problem in the artificial olfactory community. Traditionally, hardware calibration methods are laborious and ineffective due to frequent recalibration actions involving different gases, and some calibration transfer and baseline calibration methods are not effective enough. In this work, a local manifold embedding cross-domain subspace learning (LME-CDSL) model is proposed based on domain distribution alignment. It is a unified subspace learning model combined with manifold learning and domain adaptation, which tends to explore a latent transform matrix that not only enforces the drifted target domain data to learn the manifold of nondrifted source domain data but also adopts the domain adaptation method to align the domain data distribution. In general, the LME-CDSL model has three features: 1) the unsupervised and adaptation distribution subspace projection can be easily computed through eigenvector decomposition; 2) the local linear manifold learns to achieve the compact representations of high-dimensional data and is capable of preserving the local features of nondrifted samples; and 3) the domain adaptation part utilizes the maximum mean discrepancy (MMD) and variance maximization to make the sample distributions of different domains more similar and preserve the intrinsic properties. For long-term and short-term drift compensation on a single E-nose system, the local manifold embedding cross-domain subspace learning (LME-CDSL) model obtains the average recognition accuracy of 70.95% and 74.09%, respectively, while 71.71% and 73.96%, respectively for multiple identical E-nose systems with both long-term and interplate drift, which are higher than several comparative methods and proves the its effectiveness and superiority on anti-drift and gas recognition.
2 July 2020 in IEEE Access
Electronic nose (E-nose) systems have a good effect on the identification of distinct odours. However, the properties of chemical gas sensors indicate that ageing, poisoning, fluctuation of environmental conditions (moisture, temperature, etc.) and a lack of fabrication repeatability, etc. have a large impact on the sensitivity and accuracy of sensors, which leads to sensor data drift. Although previous studies have indicated the feasibility and validity of deep learning in drift compensation of gas sensor data, the actual performances of these deep learning models are less impressive compared with some existing methods. Thus, we intend to further explore a novel deep learning model for drift compensation for E-noses. In this paper, we investigate the drift compensation effect of E-nose data based on a deep belief network (DBN) and constructed a Gaussian deep belief classification network (GDBCN) model by cascading a Gaussian-Bernoulli restricted Boltzmann machines based DBN with a softmax classifier layer to compensate for sensor drift at the decision level. The merits of our method are as follows: 1) it is a unified classification model for drift auto-compensation at the decision level rather than a feature extractor; 2) it couples unsupervised and supervised techniques by modelling the intrinsic distribution of the data from different domains in an unsupervised manner and fine-tunes the model parameters by leveraging the label information of the source domain; 3) the supervised fine-tuning process for the coupled GDBCN model fits well with the nature of the supervised task and guarantees that the parameters of the DBN will be useful for classification; 4) the GDBCN model is a classification model and thus automatically compensates for drift without manually setting specific model rules for domain alignment before classification. Experimental results on real sensor datasets demonstrate the effectiveness and superiority compared with several existing control methods.
18 March 2024 in IEEE Sensors Journal
The drift compensation of gas sensor systems is an important topic in the artificial olfactory community. The drift is arisen by multiple factors, i.e., the change of temperature and moisture, poison effect, manufacture repeatability, etc. It would arise the bias, distortion and offset of data, and lead it to be unrecognized. Conventionally, the calibration methods in hardware level are laborious and ineffective, and many proposed baseline software methods are not effective enough due to loss of structure information. In this work, a classifier-level nonlinear manifold transfer extreme learning machine (NMT-ELM) is proposed, which intends to learn the manifold information embedded into the source domain data, and accomplish the structure information preservation in a nonlinear style. In addition, the domain adaptation between source and target domain is adopted as well, which is utilized for effective distribution alignment. The output matrix of extreme learning machine (ELM) is determined with supervised learning, and the generalization is enhanced effectively as well. Moreover, a novel intelligent optimization algorithm entitled Cuckoo Searching (CS) is introduced, in order to explore the optimal parameters of NMT-ELM rapidly comparing with the conventional grid searching. The validation is proved by experiments on different open-set E-nose data, and the classification accuracies comparison demonstrates its effectiveness on drift compensation of E-nose.
25 December 2023 in IEEE Transactions on Geoscience and Remote Sensing
In this study, an efficient deep-learning-driven sparse-target imaging (DLSTI) method was developed for array borehole radar to improve the accuracy of target localization in a sub-surface nonuniform media. First, by making use of the linear superposition and separable characteristics of the target and background echo, the background echo was generated with an electromagnetic simulation using prior medium information. The background echo was then removed from the radar receiver echo using a convex-optimization-based front-end target echo extractor to obtain the raw sparse target echo. Subsequently, the raw target echoes and true target locations in the simulation dataset were utilised for the training of a back-end stacked autoencoder in a data-driven manner, which is capable of illustrating accurate target locations in field tests in nonuniform environments after training. The comparison results in multiple simulations and field scenes show that the proposed DLSTI outperforms other effective imaging methods in terms of target localization accuracy and image side-lobes, including reverse time migration and back projection, whose localization error was improved to 0.02 m and the image side-lobe was reduced by 16.09 dB.
3 May 2024 in IEEE Antennas and Wireless Propagation Letters
In this letter, a waterflow velocity detection method with a 24 GHz millimetre-wave radar is presented based on a modified matrix-pencil (MP) method to alleviate the issue of stochastic waterflow surface fluctuation-induced Doppler shifts aliases in the received echo. Initially, by leveraging the echo superposition characteristics of scattering point targets, the millimetre-wave radar echo exponential polynomial model is established. Subsequently, the MP method is modified to find the low-rank approximation of the received echo for waterflow velocity estimation. The results of the 24 GHz millimetre-wave radar experiments show that the mutual difference between the modified MP-detected waterflow velocities in two azimuths is 0.036 m/s, which demonstrate the better validity and efficacy of the proposed method compared with the traditional compressive sensing and singular value decomposition methods.
28 February 2024 in IEEE Sensors Journal
In this paper, an adaptive gradient projection correction (AGPC) method for resistivity array imaging logging is proposed. Due to the irregular spin and swing of the subsurface resistivity array instrument, the time-varying nonuniform sampling is induced between the electrodes, which reduces the resolution of the wellbore image. In the proposed method, the nonuniform sampled image is reconstructed via the proposed adaptive transform determined by gradient domain minimization constraint for image correction and least-square constraint for loss minimization. The imaging correction capabilities of AGPC are theoretically simulated and experimentally demonstrated in multiple scenes. The indexes of edges correction error and average gradient difference are devised and provided to assess the AGPC performance in simulations and field experiments. The results show that AGPC achieves better balance between image loss and resolution compared with traditional methods.