Recognition Module Research Articles

Multi-branch convolutional neural network with cross-attention mechanism for emotion recognition

Research on emotion recognition is an interesting area because of its wide-ranging applications in education, marketing, and medical fields. This study proposes a multi-branch convolutional neural network model based on cross-attention mechanism (MCNN-CA) for accurate recognition of different emotions. The proposed model provides automated extraction of relevant features from multimodal data and fusion of feature maps from diverse sources as modules for the subsequent emotion recognition. In the feature extraction stage, various convolutional neural networks were designed to extract critical information from multiple dimensional features. The feature fusion module was used to enhance the inter-correlation between features based on channel-efficient attention mechanism. This innovation proves effective in fusing distinctive features within a single mode and across different modes. The model was assessed based on EEG emotion recognition experiments on the SEED and SEED-IV datasets. Furthermore, the efficiency of the proposed model was evaluated via multimodal emotion experiments using EEG and text data from the ZuCo dataset. Comparative analysis alongside contemporary studies shows that our model excels in terms of accuracy, precision, recall, and F1-score.

The Rbfox1/LASR complex controls alternative pre-mRNA splicing by recognition of multipart RNA regulatory modules.

The Rbfox proteins regulate alternative pre-mRNA splicing by binding to the RNA element GCAUG. In the nucleus, most of Rbfox is bound to the large assembly of splicing regulators (LASR), a complex of RNA-binding proteins that recognize additional RNA motifs. However, it remains unclear how the different subunits of the Rbfox/LASR complex act together to bind RNA and regulate splicing. We used a nuclease protection assay to map the transcriptome-wide footprints of Rbfox1/LASR on nascent cellular RNA. In addition to GCAUG, Rbfox1/LASR binds RNA motifs for LASR subunits hnRNPs M, H/F, and C and Matrin3. These elements are often arranged in tandem, forming multipart modules of RNA motifs. To distinguish contact sites of Rbfox1 from the LASR subunits, we analyzed a mutant Rbfox1(F125A) that has lost RNA binding but remains associated with LASR. Rbfox1(F125A)/LASR complexes no longer interact with GCAUG but retain binding to RNA elements for LASR. Splicing analyses reveal that in addition to activating exons through adjacent GCAUG elements, Rbfox can also stimulate exons near binding sites for LASR subunits. Minigene experiments demonstrate that these diverse elements produce a combined regulatory effect on a target exon. These findings illuminate how a complex of RNA-binding proteins can decode combinatorial splicing regulatory signals by recognizing groups of tandem RNA elements.

Open-Set Automatic Modulation Recognition Based on Circular Prototype Learning and Denoising Diffusion Model

Automatic modulation recognition (AMR) technology is a critical component of modern communication systems. However, conventional AMR methods based on the closed-set assumption struggle to detect unknown classes that may appear during testing. To address this limitation, this paper proposes an open-set automatic modulation recognition (OSAMR) framework, termed CPLDiff, that integrates circular prototype learning (CPL) with a denoising diffusion model (DDM) to detect unknown classes. The core idea of CPLDiff is to jointly leverage the class-level and instance-level information of the training samples. To achieve this, CPL is used to extract class-level information, while the diffusion model is employed to extract instance-level information. (1) Circular Prototype Learning: Prototype vectors are pre-optimized and fixed, and a bias radius is introduced to expand the feasible encoding space. (2) Denoising Diffusion Model: Noise is added to the sample, and the DDM is used to remove this noise. The probability of a sample belonging to a known class is proportional to the extent of noise removal. (3) Final Integration: The outputs of the CPL and the DDM are combined to perform OSAMR. We conducted comparative experiments and evaluated the proposed method using diverse metrics to ensure a comprehensive assessment of its effectiveness. The experimental results demonstrate that the CPLDiff method significantly improves the detection capability for unknown classes compared to state-of-the-art methods.

Modulation pattern recognition method of wireless communication automatic system based on IABLN algorithm in intelligent system.

The aim of this study is to address the limitations of convolutional networks in recognizing modulation patterns. These networks are unable to utilize temporal information effectively for feature extraction and modulation pattern recognition, resulting in inefficient modulation pattern recognition. To address this issue, a signal modulation recognition method based on a two-way interactive temporal attention network algorithm has been developed. A two-way interactive temporal network is designed on the basis of the long and short-term memory network with the objective of enhancing the contextual connection of the temporal network. The output of the temporal network is attentively weighted using the soft attention mechanism. The proposed algorithm exhibited enhanced overall, average, and maximum recognition rates at varying signal-to-noise ratios, with an increase of 10.34%, 8.33%, and 3.33%, respectively, in comparison to other algorithms within the Radio Machine Learning (RML) 2016.10b dataset. Furthermore, the modulated signal recognition accuracy was as high as 92.84%, with an average increase in the Kappa coefficient of 12.28%. The Kappa coefficient in the Communication Signal Processing Benchmark for Machine Learning (CSPB.ML2018) 2018 dataset was 0.62, representing an average increase of 10.32% over other algorithms. The results demonstrate that the proposed recognition method can enhance the network's accuracy in recognizing modulated signals. Moreover, it has potential applications in modulation pattern recognition in automatic systems for wireless communications.

Overview of Automatic Modulation Recognition Methods based on Deep Learning

Overview of Automatic Modulation Recognition Methods based on Deep Learning

Gesture Recognition based on Deep Learning for Quadcopters Flight Control

This article presents a system for controlling quadcopters with gestures, which are recognized by a model based on neural networks. A method based on a combined deep learning model is proposed that provides real-time recognition with minimal consumption of computing power. An implementation is presented that offers the possibility of controlling the quadcopter in two ways, via gestures or the keyboard. A functionality is also provided for adding new gestures for recognition using interactive code via the Jupyter Lab web application. A special mode is implemented that allows us to create a data set for a new test directly from the quadcopter camera to simplify data collection. The operation of the control and recognition module is demonstrated using an example in which a DJI Tello Edu drone is controlled. The results of tests under real conditions are presented. The developed software allows one to speed up the process of gesture recognition and facilitates the process of controlling the quadcopters. Several areas of improvement of the developed system and their possible technical implementation are proposed.

DNA Tetrahedron-Driven Multivalent Proteolysis-Targeting Chimeras: Enhancing Protein Degradation Efficiency and Tumor Targeting.

Proteolysis-targeting chimeras (PROTACs) are dual-functional molecules composed of a protein of interest (POI) ligand and an E3 ligase ligand connected by a linker, which can recruit POI and E3 ligases simultaneously, thereby inducing the degradation of POI and showing great potential in disease treatment. A challenge in developing PROTACs is the design of linkers and the modification of ligands to establish a multifunctional platform that enhances degradation efficiency and antitumor activity. As a programmable and modifiable nanomaterial, DNA tetrahedron can precisely assemble and selectively recognize molecules and flexibly adjust the distance between molecules, making them ideal linkers. Herein, we developed a multivalent PROTAC based on a DNA tetrahedron, named AS-TD2-PRO. Using DNA tetrahedron as a linker, we combined modules targeting tumor cells, recognizing E3 ligases, and multiple POI together. We took the undruggable target protein signal transducer and activator of transcription 3 (STAT3), associated with the etiology and progression in a variety of malignant tumors, as an example in this study. AS-TD2-PRO with two STAT3 recognition modules demonstrated good potential in enhancing tumor-specific targeting and degradation efficiency compared to traditional bivalent PROTACs. Furthermore, in a mouse tumor model, the superior therapeutic activity of AS-TD2-PRO was observed. Overall, DNA tetrahedron-driven multivalent PROTACs both serve as a proof of principle for multifunctional PROTAC design and introduce a promising avenue for cancer treatment strategies.

Research on automatic modulation recognition of shortwave signals in few‐shot scenarios based on knowledge distillation

AbstractAutomatic modulation recognition plays an important role in wireless communication and radio regulation. Existing deep learning‐based automatic modulation recognition techniques perform well with large datasets and high computational power but require significant resources for labelled data and complex pre‐processing. This paper proposes a multi‐information fusion method for few‐shot modulation recognition, which involves converting I/Q signals into A/P signals and training with a combination of VGG and LSTM network models. An ensemble knowledge distillation (EKD) approach is employed to streamline the network model, meeting the demands for deploying neural network models on compact devices. Experimental results demonstrate that using only 1% of the shortwave modulation signal dataset as the training set, the proposed model achieves an average classification accuracy of 71.08% under all signal‐to‐noise ratios, surpassing the currently popular deep learning models. Moreover, two small‐scale networks, MobileNetV3 and convolutional neural network are trained, through EKD. Compared to the teacher network, the floating‐point operations of the distilled models are reduced by 99.8% and 99.7%, respectively, and the average prediction accuracy only decreases by 16.05% and 8.09%. The lightweight, few‐shot networks designed in this study for shortwave modulation signals aim to achieve fast and accurate modulation recognition on compact devices.

Multimodal Control for Powered Wheelchair

Aim: Over the years, joystick-controlled wheelchairs have helped in the mobility of non-ambulant people. However, existing wheelchairs may not be completely helpful for people suffering from additional medical conditions such as, dysarthria, cerebral palsy, and diplegic users. The current work aims to develop a speech and gesture multimodal controlled wheelchair system for people with multiple disabilities. The motivation behind this work is to cater to their needs by customizing their wheelchairs for ease of movement and to instill a feeling of independence when they commute. Methods: The speech recognition module of the proposed wheelchair is developed to cater the needs of both normal and dysarthric speakers. The current work makes use of convolution and residual convolution based neural architectures to model and recognise the speech commands. Further, a range-based gesture recognition system is developed in the current work using MPU 6050. Results: The proposed speech recognition module of the wheelchair has an accuracy up to 98% and 92% for normal and dysarthric speakers respectively. Further the gesture recognition module has an accuracy of 100% for all users as it is completely threshold based. Conclusion: The proposed speech module is ported on to a Raspberry pi 4B+ module and the gesture module is ported using an Arduino Uno. These modules are integrated with the control system (i.e., joystick) of the wheelchair to ensure movement. The current work ensures a small footprint and easy mobility of the wheelchair.

A multi-task learning framework for joint sub-Nyquist wideband spectrum sensing and modulation recognition

A multi-task learning framework for joint sub-Nyquist wideband spectrum sensing and modulation recognition

A Pre-Activation Residual Convolutional Network With Attention Modules for High-Resolution Segmented EEG Emotion Recognition

A Pre-Activation Residual Convolutional Network With Attention Modules for High-Resolution Segmented EEG Emotion Recognition

Secure and Intelligent Sensing in Unmanned Aerial Vehicles: A Semi-supervised Modulation Recognition Framework

Secure and Intelligent Sensing in Unmanned Aerial Vehicles: A Semi-supervised Modulation Recognition Framework

EEG-based emotion recognition using multi-scale dynamic CNN and gated transformer

Emotions play a crucial role in human thoughts, cognitive processes, and decision-making. EEG has become a widely utilized tool in emotion recognition due to its high temporal resolution, real-time monitoring capabilities, portability, and cost-effectiveness. In this paper, we propose a novel end-to-end emotion recognition method from EEG signals, called MSDCGTNet, which is based on the Multi-Scale Dynamic 1D CNN and the Gated Transformer. First, the Multi-Scale Dynamic CNN is used to extract complex spatial and spectral features from raw EEG signals, which not only avoids information loss but also reduces computational costs associated with the time-frequency conversion of signals. Then, the Gated Transformer Encoder is utilized to capture global dependencies of EEG signals. This encoder focuses on specific regions of the input sequence while reducing computational resources through parallel processing with the improved multi-head self-attention mechanisms. Third, the Temporal Convolution Network is used to extract temporal features from the EEG signals. Finally, the extracted abstract features are fed into a classification module for emotion recognition. The proposed method was evaluated on three publicly available datasets: DEAP, SEED, and SEED_IV. Experimental results demonstrate the high accuracy and efficiency of the proposed method for emotion recognition. This approach proves to be robust and suitable for various practical applications. By addressing challenges posed by existing methods, the proposed method provides a valuable and effective solution for the field of Brain-Computer Interface (BCI).

Practical Design of a Home Appliance Control System Using Audio Signals

In this paper, the design of a system that controls home appliances using voice commands is considered. This system may be utilized by people with special needs and the elderly. By using just their voice, they can control the lighting of their room, and turn on or off their television, air conditioner, or fan with safety and security and without needing someone else's assistance. This may be achieved by passing their voice commands to a microcontroller, (Arduino), through a device that uses voice or speech recognition technology. Herein, the system is designed at the lowest possible costs. Two basic ways are discussed to build this system, a wired system that does not depend on wireless technologies and the other depends on Bluetooth technology. Each of them has advantages and disadvantages. The hardware parts utilized in the design process are the Arduino Uno, Relay Module, and voice recognition module. The latter could be replaced by a Bluetooth module and a smartphone with an application that uses voice or speech recognition technology. The proposed wired system proved to be advantageous in terms of privacy, security and reliability. A comparison between the two systems is concluded at the end of this paper.

Analysis on dendritic deep learning model for AMR task

This study introduces a novel hybrid deep learning model featuring a dendritic layer for enhancing the performance of automatic modulation recognition (AMR). By replacing the fully connected layer, the proposed model demonstrates superior classification accuracy in AMR tasks. Comparative experiments with nine state-of-the-art deep learning models on the RadioML2016.10a dataset reveal its consistent superiority. Statistical analyses, including the Friedman test and Wilcoxon signed-rank test, confirm the significant advantage of the HDM-D model.

Signal Modulation Recognition via Bias Adjustment-Based Class Incremental Learning

Signal Modulation Recognition via Bias Adjustment-Based Class Incremental Learning

Enhancing Gesture-Based Interactions for Individuals with Disabilities through Dual-Attention Wireless Sensing Networks

This paper investigates the application of wireless sensing technology to develop a more natural and intuitive gesture-based interaction framework for individuals with disabilities. The study begins by identifying the primary challenges faced by this demographic when interacting with conventional devices. It then introduces a novel approach for WiFi-enabled gesture recognition and interpretation using dual attention networks. This methodology comprises two essential components: Channel State Information (CSI) preprocessing and a gesture recognition module. The paper details the implementation of these modules and elucidates their roles in enhancing gesture detection accuracy. Furthermore, the discussion extends to the future prospects of wireless sensing technologies, envisioning their integration into smart home systems, public services, and enhanced social interactions. The research underscores the transformative potential of integrating gesture-based interactions with wireless sensing to significantly elevate the quality of life for people with disabilities, suggesting a paradigm shift in how assistive technologies are developed and utilized.

A lightweight model for echo trace detection in echograms based on improved YOLOv8

With the rise of underwater unmanned platforms like unmanned boats, ROVs, and AUVs, there’s an increasing need for underwater detection technologies. Researchers have merged scientific echosounders with these platforms for biometric applications. However, current detection models are too parameter-heavy to embed in echosounders and struggle with noisy, irregular, and dense echograms. This paper introduces YOLOv8-SBE, a lightweight fish detection model based on YOLOv8, addressing these issues by enhancing feature extraction, information fusion, and small object recognition. YOLOv8-SBE adds the C2f_ScConv module to improve efficiency and reduce parameters, incorporates the BiFPN structure to enhance information transfer, and uses the EMA attention module for better small target recognition. It reduces computational complexity by 18.5%, decreases model parameters by 40%, and improves mAP0.5 to 79.5% and mAP0.5:0.95 to 58.2%, making it suitable for echosounders with limited resources.

An Automatic Modulation Recognition Algorithm Based on Time–Frequency Features and Deep Learning with Fading Channels

Automatic modulation recognition (AMR) stands as a crucial core technology within the realm of signal processing and perception, playing a significant part in harsh electromagnetic environments. The time–frequency image (TFI) of communication signals can manifest modulation characteristics and serve as a foundation for signal modulation recognition and classification. However, under the influence of the electromagnetic environment, communication signals are exposed to varying degrees of interference, which poses a challenge to the recognition of modulation types. Taking into account the effects of interference and channel fading, this paper introduces a communication signal modulation recognition algorithm based on deep learning (DL) and time–frequency analysis. This approach employs short-time Fourier transform (STFT) to generate time–frequency diagrams from time-domain signals. Subsequently, it binarizes the image and feeds it as input data to the neural network. Our research presents a composite deep convolutional neural network (CNN) architecture known as the composite dense-residual neural network (CDRNN). This architecture focuses on enhancing the feature extraction and identification, aiming to achieve accurate recognition of modulation types in harsh electromagnetic environments. Finally, simulation results validate that the proposed deep learning algorithm holds remarkable advantages in boosting the accuracy of modulation type recognition with better adaptability. The algorithm shows better performance even in harsh electromagnetic environments. When the signal-to-noise ratio (SNR) is 18 dB, the recognition accuracy can reach 92.1%.

A transfer learning-based GAN for data augmentation in automatic modulation recognition

Abstract The automatic modulation recognition (AMR) algorithms based on deep learning (DL) have achieved high classification accuracy by automatically extracting deep features from massive data. However, in real-world scenarios, sufficient training data is always difficult to collect, which affects the performance of DL-based models. As a type of data augmentation algorithm, Generative Adversarial Networks (GANs) can generate artificial data similar to the given real data and thus solve the problem of insufficient data, whereas the training process of GANs is also affected by limited number of data samples. Inspired by the successful application of transferring GANs in the field of image generation, this paper employs transfer learning-based GANs to enlarge the training data by generating the constellation diagram images of radio signals, which can effectively solve the problems of divergence and model collapse. We feed the enhanced dataset into a CNN model for modulation recognition and the experimental results demonstrate that our proposed method achieves a performance improvement ranging from 1% to 5.1% compared with the result of the original limited training data.