Hand-crafted Feature Engineering Research Articles

Automated crack localization for road safety using contextual u-net with spatial-channel feature integration

Accurate and timely crack localization is crucial for road safety and maintenance, but image processing and hand-crafted feature engineering methods, often fail to distinguish cracks from background noise under diverse lighting and surface conditions. This paper proposes a framework utilizing contextual U-Net deep learning model to automatically localize cracks in road images. The framework design considers crack localization as a task of pixel-level segmenting, and analyzing each pixel in a road image to determine if it belongs to a crack or not. The proposed U-Net model uses a robust EfficientNet encoder to capture crucial details (spatial features) and overall patterns (channel-wise features) within the road image. This balanced approach helps the model learn effectively from both individual elements and the context of the images, leading to improved crack detection. A customized hierarchical attention mechanism is designed to make U-Net model contextually adaptive to focus on relevant features at different scales and resolutions for accurately localizing road cracks that can vary widely in size and shape. The model's effectiveness is demonstrated through extensive evaluations on the benchmarked and custom-made datasets.

A CNN Approach for Emotion Recognition via EEG

Emotion recognition via electroencephalography (EEG) has been gaining increasing attention in applications such as human–computer interaction, mental health assessment, and affective computing. However, it poses several challenges, primarily stemming from the complex and noisy nature of EEG signals. Commonly adopted strategies involve feature extraction and machine learning techniques, which often struggle to capture intricate emotional nuances and may require extensive handcrafted feature engineering. To address these limitations, we propose a novel approach utilizing convolutional neural networks (CNNs) for EEG emotion recognition. Unlike traditional methods, our CNN-based approach learns discriminative cues directly from raw EEG signals, bypassing the need for intricate feature engineering. This approach not only simplifies the preprocessing pipeline but also allows for the extraction of more informative features. We achieve state-of-the-art performance on benchmark emotion datasets, namely DEAP and SEED datasets, showcasing the superiority of our approach in capturing subtle emotional cues. In particular, accuracies of 96.32% and 92.54% were achieved on SEED and DEAP datasets, respectively. Further, our pipeline is robust against noise and artefact interference, enhancing its applicability in real-world scenarios.

AttenSyn: An Attention-Based Deep Graph Neural Network for Anticancer Synergistic Drug Combination Prediction.

Identifying synergistic drug combinations is fundamentally important to treat a variety of complex diseases while avoiding severe adverse drug-drug interactions. Although several computational methods have been proposed, they highly rely on handcrafted feature engineering and cannot learn better interactive information between drug pairs, easily resulting in relatively low performance. Recently, deep-learning methods, especially graph neural networks, have been widely developed in this area and demonstrated their ability to address complex biological problems. In this study, we proposed AttenSyn, an attention-based deep graph neural network for accurately predicting synergistic drug combinations. In particular, we adopted a graph neural network module to extract high-latent features based on the molecular graphs only and exploited the attention-based pooling module to learn interactive information between drug pairs to strengthen the representations of drug pairs. Comparative results on the benchmark datasets demonstrated that our AttenSyn performs better than the state-of-the-art methods in the prediction of anticancer synergistic drug combinations. Additionally, to provide good interpretability of our model, we explored and visualized some crucial substructures in drugs through attention mechanisms. Furthermore, we also verified the effectiveness of our proposed AttenSyn on two cell lines by visualizing the features of drug combinations learnt from our model, exhibiting satisfactory generalization ability.

Understanding the class separation process of convolutional neural networks

Extensive research shows that deep neural networks (DNNs) capture better feature representations than the previous handcrafted feature engineering, which leads to a significant performance improvement. However, it raises the question of how does the neural network spontaneously learn the intermediate representation, during the training process, to correctly separate the test data set into different categories? In this paper, based on the existing research work, we continue to take a small step towards understanding the dynamics of convolutional neural networks (CNN). Specifically, we model the test data set as a relationship graph, and each intermediate layer representation learned by CNN will transform the relationship graph correspondingly. Through these continuous transformations, we investigate the class separation process of CNN from two perspectives: 1) show the evolution of the degree of correlation between samples through the histogram of similarity between vertices; 2) by visualizing and quantifying the degree of separation (i.e., modularity) of the important relationship subgraph, we can show the contribution of each convolution layer of CNN to class separation. In the preliminary experiment, it can be found that: 1) each convolution layer of CNN gradually reduced the similarity of most edges between the samples to a very low level, that is, the dense relationship graph gradually evolved into a sparse graph; 2) the remaining important relationship subgraph consist of edges with high similarity shows evident modularity; 3) both visualization and quantification results show that the modularity of intra-class subgraph increases when the layers go deeper. Moreover, the degradation and plateau in the modularity curve reveal the existence of redundant layers. In this paper, we reveal some more in-depth dynamics of CNN, and introduced another modularity, which is widely used as the theoretical guidance tool of layer pruning. It can save more model parameters without losing the accuracy of classification.

CACPP: A Contrastive Learning-Based Siamese Network to Identify Anticancer Peptides Based on Sequence Only.

Anticancer peptides (ACPs) recently have been receiving increasing attention in cancer therapy due to their low consumption, few adverse side effects, and easy accessibility. However, it remains a great challenge to identify anticancer peptides via experimental approaches, requiring expensive and time-consuming experimental studies. In addition, traditional machine-learning-based methods are proposed for ACP prediction mainly depending on hand-crafted feature engineering, which normally achieves low prediction performance. In this study, we propose CACPP (Contrastive ACP Predictor), a deep learning framework based on the convolutional neural network (CNN) and contrastive learning for accurately predicting anticancer peptides. In particular, we introduce the TextCNN model to extract the high-latent features based on the peptide sequences only and exploit the contrastive learning module to learn more distinguishable feature representations to make better predictions. Comparative results on the benchmark data sets indicate that CACPP outperforms all the state-of-the-art methods in the prediction of anticancer peptides. Moreover, to intuitively show that our model has good classification ability, we visualize the dimension reduction of the features from our model and explore the relationship between ACP sequences and anticancer functions. Furthermore, we also discuss the influence of data set construction on model prediction and explore our model performance on the data sets with verified negative samples.

ExamPle: explainable deep learning framework for the prediction of plant small secreted peptides.

Plant Small Secreted Peptides (SSPs) play an important role in plant growth, development, and plant-microbe interactions. Therefore, the identification of SSPs is essential for revealing the functional mechanisms. Over the last few decades, machine learning-based methods have been developed, accelerating the discovery of SSPs to some extent. However, existing methods highly depend on handcrafted feature engineering, which easily ignores the latent feature representations and impacts the predictive performance. Here, we propose ExamPle, a novel deep learning model using Siamese network and multi-view representation for the explainable prediction of the plant SSPs. Benchmarking comparison results show that our ExamPle performs significantly better than existing methods in the prediction of plant SSPs. Also, our model shows excellent feature extraction ability. Importantly, by utilizing in silicomutagenesis experiment, ExamPle can discover sequential characteristics and identify the contribution of each amino acid for the predictions. The key novel principle learned by our model is that the head region of the peptide and some specific sequential patterns are strongly associated with the SSPs' functions. Thus, ExamPle is expected to be a useful tool for predicting plant SSPs and designing effective plant SSPs. Our codes and datasets are available at https://github.com/Johnsunnn/ExamPle.

Capacity estimation of lithium-ion batteries based on data aggregation and feature fusion via graph neural network

Lithium-ion batteries in electrical devices face inevitable degradation along with the long-term usage. The accompanying battery capacity estimation is crucial for battery health management. However, the hand-crafted feature engineering in traditional methods and complicated network design followed by the laborious trial in data-driven methods hinder efficient capacity estimation. In this work, the battery measurements from different sensors are organized as the graph structure and comprehensively utilized based on graph neural network. The feature fusion is further designed to enhance the network capacity. The specific data aggregation and feature fusion operations are selected by neural architecture search, which relieves the network design and increases the adaptability. Two public datasets are adopted to verify the effectiveness of the proposed scheme. Additional discussions are conducted to emphasize the capability of the graph neural network and the necessity of architecture searching. The comparison analysis and the performance under noisy environment further demonstrate the superiority of proposed scheme.

Spikebench: An open benchmark for spike train time-series classification

Modern well-performing approaches to neural decoding are based on machine learning models such as decision tree ensembles and deep neural networks. The wide range of algorithms that can be utilized to learn from neural spike trains, which are essentially time-series data, results in the need for diverse and challenging benchmarks for neural decoding, similar to the ones in the fields of computer vision and natural language processing. In this work, we propose a spike train classification benchmark, based on open-access neural activity datasets and consisting of several learning tasks such as stimulus type classification, animal's behavioral state prediction, and neuron type identification. We demonstrate that an approach based on hand-crafted time-series feature engineering establishes a strong baseline performing on par with state-of-the-art deep learning-based models for neural decoding. We release the code allowing to reproduce the reported results.

Seizure onset zone identification using phase-amplitude coupling and multiple machine learning approaches for interictal electrocorticogram

Automatic seizure onset zone (SOZ) localization using interictal electrocorticogram (ECoG) improves the diagnosis and treatment of patients with medically refractory epilepsy. This study aimed to investigate the characteristics of phase-amplitude coupling (PAC) extracted from interictal ECoG and the feasibility of PAC serving as a promising biomarker for SOZ identification. We employed the mean vector length modulation index approach on the 20-s ECoG window to calculate PAC features between low-frequency rhythms (0.5–24 Hz) and high frequency oscillations (HFOs) (80–560 Hz). We used statistical measures to test the significant difference in PAC between the SOZ and non-seizure onset zone (NSOZ). To overcome the drawback of handcraft feature engineering, we established novel machine learning models to learn automatically the characteristics of the obtained PAC features and classify them to identify the SOZ. Besides, to handle imbalanced dataset classification, we introduced novel feature-wise/class-wise re-weighting strategies in conjunction with classifiers. In addition, we proposed a time-series nest cross-validation to provide more accurate and unbiased evaluations for this model. Seven patients with focal cortical dysplasia were included in this study. The experiment results not only showed that a significant coupling at band pairs of slow waves and HFOs exists in the SOZ when compared with the NSOZ, but also indicated the effectiveness of the PAC features and the proposed models in achieving better classification performance .

On the combination of graph data for assessing thin-file borrowers’ creditworthiness

Thin-file borrowers are customers for whom a creditworthiness assessment is uncertain due to their lack of credit history. To address missing credit information, many researchers have used borrowers’ social interactions as an alternative data source. Exploiting social networking data has traditionally been achieved by hand-crafted feature engineering, but lately, graph neural networks have emerged as a promising alternative. Here we introduce an information-processing framework to improve credit scoring models by blending several methods of graph representation learning: feature engineering, graph embeddings, and graph neural networks. In this approach, we aggregate the methods’ outputs to be fed to a gradient boosting classifier to produce a final creditworthiness score. We have validated this framework over a unique multi-source dataset that characterizes the relationships, interactions, and credit history for the entire population of a Latin American country, applying it to credit risk models, application, and behavior. It also allows us to study both individuals and companies. Our results show that the methods of graph representation learning should be used as complements; they should not be seen as self-sufficient methods, as it is currently done. We improve the creditworthiness assessment performance in terms of the measures of Area Under the ROC Curve (AUC) and Kolmogorov–Smirnov (KS), outperforming traditional methods of exploiting social interaction data. In the area of corporate lending, where the potential gain is much higher, our results confirm that the evaluation of a thin-file company cannot solely consider the company’s own characteristics. The business ecosystem in which these companies interact with their owners, suppliers, customers, and other companies provides novel knowledge that enables financial institutions to enhance their creditworthiness assessment. Our results let us know when and on which population to use graph data and the expected effects on performance. They also show the enormous value of graph data on the credit scoring problem for thin-file borrowers, mainly to help companies with thin or no credit history to enter the financial system.

Op2Vec: An Opcode Embedding Technique and Dataset Design for End-to-End Detection of Android Malware

Android is one of the leading operating systems for smartphones in terms of market share and usage. Unfortunately, it is also an appealing target for attackers to compromise its security through malicious applications. To tackle this issue, domain experts and researchers are trying different techniques to stop such attacks. All the attempts of securing the Android platform are somewhat successful. However, existing detection techniques have severe shortcomings, including the cumbersome process of feature engineering. Designing representative features require expert domain knowledge. There is a need for minimizing human experts’ intervention by circumventing handcrafted feature engineering. Deep learning could be exploited by extracting deep features automatically. Previous work has shown that operational codes (opcodes) of executables provide key information to be used with deep learning models for the detection process of malicious applications. The only challenge is to feed opcodes information to deep learning models. Existing techniques use one-hot encoding to tackle the challenge. However, the one-hot encoding scheme has severe limitations. In this paper, we introduce (1) a novel technique for opcodes embedding, which we name Op2Vec, and (2) based on the learned Op2Vec, we have developed a dataset for end-to-end detection of Android malware. Introducing the end-to-end Android malware detection technique avoids expert-intensive handcrafted feature extraction and ensures automation. Some of the recent deep learning-based techniques showed significantly improved results when tested with the proposed approach and achieved an average detection accuracy of 97.47%, precision of 0.976, and F1 score of 0.979.

An Accurate Multiple Sclerosis Detection Model Based on Exemplar Multiple Parameters Local Phase Quantization: ExMPLPQ

Multiple sclerosis (MS) is a chronic demyelinating condition characterized by plaques in the white matter of the central nervous system that can be detected using magnetic resonance imaging (MRI). Many deep learning models for automated MS detection based on MRI have been presented in the literature. We developed a computationally lightweight machine learning model for MS diagnosis using a novel handcrafted feature engineering approach. The study dataset comprised axial and sagittal brain MRI images that were prospectively acquired from 72 MS and 59 healthy subjects who attended the Ozal University Medical Faculty in 2021. The dataset was divided into three study subsets: axial images only (n = 1652), sagittal images only (n = 1775), and combined axial and sagittal images (n = 3427) of both MS and healthy classes. All images were resized to 224 × 224. Subsequently, the features were generated with a fixed-size patch-based (exemplar) feature extraction model based on local phase quantization (LPQ) with three-parameter settings. The resulting exemplar multiple parameters LPQ (ExMPLPQ) features were concatenated to form a large final feature vector. The top discriminative features were selected using iterative neighborhood component analysis (INCA). Finally, a k-nearest neighbor (kNN) algorithm, Fine kNN, was deployed to perform binary classification of the brain images into MS vs. healthy classes. The ExMPLPQ-based model attained 98.37%, 97.75%, and 98.22% binary classification accuracy rates for axial, sagittal, and hybrid datasets, respectively, using Fine kNN with 10-fold cross-validation. Furthermore, our model outperformed 19 established pre-trained deep learning models that were trained and tested with the same data. Unlike deep models, the ExMPLPQ-based model is computationally lightweight yet highly accurate. It has the potential to be implemented as an automated diagnostic tool to screen brain MRIs for white matter lesions in suspected MS patients.

Identifying Adverse Drug Reaction-Related Text from Social Media: A Multi-View Active Learning Approach with Various Document Representations

Adverse drug reactions (ADRs) are a huge public health issue. Identifying text that mentions ADRs from a large volume of social media data is important. However, we need to address two challenges for high-performing ADR-related text detection: the data imbalance problem and the requirement of simultaneously using data-driven information and handcrafted information. Therefore, we propose an approach named multi-view active learning using domain-specific and data-driven document representations (MVAL4D), endeavoring to enhance the predictive capability and alleviate the requirement of labeled data. Specifically, a new view-generation mechanism is proposed to generate multiple views by simultaneously exploiting various document representations obtained using handcrafted feature engineering and by performing deep learning methods. Moreover, different from previous active learning studies in which all instances are chosen using the same selection criterion, MVAL4D adopts different criteria (i.e., confidence and informativeness) to select potentially positive instances and potentially negative instances for manual annotation. The experimental results verify the effectiveness of MVAL4D. The proposed approach can be generalized to many other text classification tasks. Moreover, it can offer a solid foundation for the ADR mention extraction task, and improve the feasibility of monitoring drug safety using social media data.

ABL-TC: A lightweight design for network traffic classification empowered by deep learning

Network traffic classification is an increasingly significant prerequisite for network management. An accurate traffic classifier can contribute to traffic engineering, traffic intrusion detection and user behavior analysis. Recently, deep learning (DL) has attracted considerable attention for traffic classification due to its excellent learning ability without the need for handcrafted feature engineering or privacy invasion. However, most DL-based traffic classification solutions mainly focus on improving the identification performance, with no regard for the computational and memory burden caused by a large model size or data redundancy. Motivated by the above gaps, we elaborate on the preprocessing of raw traffic data and investigate a lightweight traffic classifier. Specifically, we design a preprocessing approach to convert raw traffic data into available datasets for deep learning based traffic classifiers, which tailors raw traffic data as training datasets by resolving its structure and content and pruning redundant information. Thanks to its ability to focus on key information, attention mechanism is introduced to design an attention-based long short-term memory (LSTM) model for traffic classification, termed the ABL-TC. ABL-TC can effectively identify global dependencies between the input and the output with a limited number of important raw features, which can contribute to learning key information to identify various traffic types. Extensive experiments on real-world public datasets show that the ABL-TC outperforms state-of-the-art approaches in terms of various metrics for recognizing traffic categories while remaining competitive in terms of time and memory efficiency.

A Demand-Side Load Event Detection Algorithm Based on Wide-Deep Neural Networks and Randomized Sparse Backpropagation

Event detection is an important application in demand-side management. Precise event detection algorithms can improve the accuracy of non-intrusive load monitoring (NILM) and energy disaggregation models. Existing event detection algorithms can be divided into four categories: rule-based, statistics-based, conventional machine learning, and deep learning. The rule-based approach entails hand-crafted feature engineering and carefully calibrated thresholds; the accuracies of statistics-based and conventional machine learning methods are inferior to the deep learning algorithms due to their limited ability to extract complex features. Deep learning models require a long training time and are hard to interpret. This paper proposes a novel algorithm for load event detection in smart homes based on wide and deep learning that combines the convolutional neural network (CNN) and the soft-max regression (SMR). The deep model extracts the power time series patterns and the wide model utilizes the percentile information of the power time series. A randomized sparse backpropagation (RSB) algorithm for weight filters is proposed to improve the robustness of the standard wide-deep model. Compared to the standard wide-deep, pure CNN, and SMR models, the hybrid wide-deep model powered by RSB demonstrates its superiority in terms of accuracy, convergence speed, and robustness.

A novel deep recurrent convolutional neural network for subthalamic nucleus localization using local field potential signals

Deep brain stimulation (DBS) of the subthalamic nucleus (STN) is a well-established interventional treatment for improving motor symptoms of patients suffering from Parkinson’s disease (PD). While STN is originally localized using imaging modalities, additional intraoperative guidance such as microelectrode recording (MER) is crucial to refine the final electrode trajectory. Analysis of MER by an experienced neurophysiologist maintains good clinical outcomes, although the procedure requires long duration and jeopardizes the safety of the surgery. Lately, local field potentials (LFP) investigation has inspired the emergence of adaptive DBS and revealed beneficial perception of PD mechanisms. Several studies confronting LFP analysis to detect the anatomical borders of STN, have focused on handcrafted feature engineering, which does not certainly capture delicate differences in LFP. This study gauges the ability of deep learning to exhibit valuable insight into the electrophysiological neural rhythms of STN using LFP. A recurrent convolutional neural network (CNN) strategy is presented, where local features are extracted from LFP signals via CNN, followed by recurrent layers to aggregate the best features for classification. The proposed model outperformed the state-of-the-art techniques, yielding highest average accuracy of 96.79%. This is the first study on the analysis of LFP signals to localize STN using deep recurrent CNN. The developed model has the potential to extract high level biomarkers regarding STN region, which would boost the neurosurgeon’s confidence in adjusting the trajectory intraoperatively for optimal lead implantation. LFP is a robust guidance tool and could be an alternative solution to the current scenario using MER.

A Multilevel Convolutional Recurrent Neural Network for Blade Icing Detection of Wind Turbine

Blade icing detection becomes increasingly significant as it can avoid revenue loss and power degradation. Conventional methods are usually limited by additional costs, and model-driven methods heavily depend on prior domain knowledge. Data-driven methods, especially deep learning approaches without needing the time-consuming handcraft feature engineering, offer a promising solution for blade icing detection. However, the monitoring signals normally have complex and diverse features as wind turbine operates in complex environments, thus effective model is needed for data analyzing. Additionally, the distribution of monitoring data is imbalanced, which causes the abnormal data mining inadequate. In this work, a multilevel convolutional recurrent neural network (MCRNN), is proposed for blade icing detection. Specifically, discrete wavelet decomposition is leveraged to obtain multilevel features both from the time domain and the frequency domain. A parallel structure combining an LSTM branch and a CNN branch is established in each level for feature extraction. To alleviate the severe data imbalance, two mechanisms, including data resampling algorithm and class-rebalanced loss function, are investigated. Furthermore, a multi-step accumulation strategy is proposed to enhance the accuracy of real-time detection. Extensive studies demonstrate that the proposed MCRNN can achieve up 38.8% and 42.9% higher F1-score over the best baseline on the balanced data sets processed by data resampling algorithm and 23.9% and 30.6% higher on imbalanced data sets with MCRNN optimized by the class-rebalanced loss function. The real-time detection verifies the applicability of the proposed method and indicates that the proposed multi-step accumulation strategy can improve the accuracy of icing detection.

ATCSpeechNet: A multilingual end-to-end speech recognition framework for air traffic control systems

In this paper, a multilingual end-to-end framework, called ATCSpeechNet, is proposed to tackle the issue of translating communication speech into human-readable text in air traffic control (ATC) systems. In the proposed framework, we focus on integrating multilingual automatic speech recognition (ASR) into one model, in which an end-to-end paradigm is developed to convert speech waveforms into text directly, without any feature engineering or lexicon. To compensate the deficiency of handcrafted feature engineering caused by ATC challenges, including multilingual, multispeaker dialog and unstable speech rates, a speech representation learning (SRL) network is proposed to capture robust and discriminative speech representations from raw waves. The self-supervised training strategy is adopted to optimize the SRL network from unlabeled data, and to further predict the speech features, i.e., wave-to-feature. An end-to-end architecture is improved to complete the ASR task, in which a grapheme-based modeling unit is applied to address the multilingual ASR issue. Facing the problem of small transcribed samples in the ATC domain, an unsupervised approach with mask prediction is applied to pretrain the backbone network of the ASR model on unlabeled data by a feature-to-feature process. Finally, by integrating the SRL with ASR, an end-to-end multilingual ASR framework is formulated in a supervised manner, which is able to translate the raw wave into text in one model, i.e., wave-to-text. Experimental results on the ATCSpeech corpus demonstrate that the proposed approach achieves high performance with a very small labeled corpus and less resource consumption, only a 4.20% label error rate on the 58-hour transcribed corpus. Compared to the baseline model, the proposed approach obtains over 100% relative performance improvement which can be further enhanced with increasing size of the transcribed samples. It is also confirmed that the proposed SRL and training strategies make significant contributions to improving the final performance. In addition, the effectiveness of the proposed framework is also validated on common corpora (AISHELL, LibriSpeech, and cv-fr). More importantly, the proposed multilingual framework not only reduces the system complexity but also obtains higher accuracy compared to that of the independent monolingual ASR models. The proposed approach can also greatly reduce the cost of annotating samples, which benefits to advance the ASR technique to industrial applications.

Risk assessment for contagion path in complex loan network

Small and medium-sized enterprises (SMEs) loans play an essential role in many aspects: including technological innovation, economic development, employment, and peoples livelihood, etc. In order to meet the loan evaluation criteria of commercial banks, many SMEs choose to guarantee each other to obtain loans, thus forming a complex guarantee network. If the borrower defaults on the loan, the risk will be diffused to its guarantors along with the contagion path, which may lead to systemic risk across the loan networks. This has brought severe challenges to the nations financial security and regulation. Thus, accurately rating the contagion path is an urgent task for systematic risk management in the loan network. Therefore, we present a deep learning-based approach to the risk rating of contagion paths in the bank industry. We leverage the graph neural network and attention mechanism on graph-structured loan behavior to learn high-order representations, which do not require handcraft feature engineering. We demonstrate that our approach outperforms the existing baselines with 2%$\\sim$15% improvements in risk rating and 3.5% in the newly constructed path rating problem. The result demonstrates the effectiveness of our proposed approach, which provides an effective method and theory basis for regulatory commissions and financial institutions to monitor systematic risks in networked-guarantee loans.

Analysis of epileptic seizures based on EEG using recurrence plot images and deep learning

This work proposes deep learning (DL) based epileptic seizure detection by generating 2D recurrence plot (RP) images of EEG signals for specific brain rhythms. The DL bypasses hand-crafted feature engineering, but extracts feature automatically from input images has displayed significant performance in various domain classification tasks. However, generating 2D images from 1D EEG signals and its quality assessment for DL pipeline has not been addressed properly, which is very crucial as the performance of the DL highly relies on input quality. Besides, suitable brain rhythm for seizure analysis has not been explored properly. Therefore, in this work, 2D input images have been generated by the RP technique from EEG signals for specific brain rhythms by preserving the nonlinear characteristics of EEG and employed a well-known DL, called convolution neural network (CNN). For, experimental validation, two well recognized EEG databases for seizure analysis from Bonn University and CHB-MIT (PhysioNet) have been considered. Eventually, three major parameters — recurrence threshold, time delay, and embedding dimension for an RP image generation have been evaluated and detailed. The results show that the proposed method can achieve classification accuracy up to 93%, which is significantly higher and the δ rhythm has been found suitable for seizure detection. The entropy of RP has been found as a suitable parameter for image quality assessment along with two global statistical parameters such as skewness of root mean square and standard of RP images. In performance evaluation, the proposed method demonstrates its competency by displaying the best classification accuracy compared to related works.