Perceptual Mechanisms Research Articles

A novel multi-sensor local and global feature fusion architecture based on multi-sensor sparse Transformer for intelligent fault diagnosis

Deep learning has been widely used for intelligent fault diagnosis of rotating machinery. However, owing to the limitations of training sample data and the complex industrial environments with variable operating conditions and noise interference, the existing deep learning-based fault diagnosis methods have difficulty achieving satisfactory performance. To address these issues, this paper proposes a novel multi-sensor local and global feature fusion architecture for intelligent fault diagnosis. First, through the integration of a stem structure and one-dimensional convolutions, a local feature perception mechanism is constructed to learn the sensor-specific local features within each sensor. Second, a global feature perception mechanism, which incorporates a multi-sensor sparse Transformer and a hierarchical architecture, is established to fully explore the sensor-specific global features within each sensor and the cross-sensor global features among multiple sensors. Third, the sensor-specific and cross-sensor global features are fed into a feature aggregation module to obtain the final multi-sensor global features. Finally, these multi-sensor global features are classified through a multi-sensor feature classifier to obtain diagnostic results. The experimental results obtained for a gear case and an inter-shaft bearing case demonstrate the superior diagnostic performance of the proposed method compared to the state-of-the-art comparative methods under limited training samples and complex environments.

Supplemental Material for A Perceptual Cue-Based Mechanism for Automatic Assignment of Thematic Agent and Patient Roles

Supplemental Material for A Perceptual Cue-Based Mechanism for Automatic Assignment of Thematic Agent and Patient Roles

Pain characteristics in patients with hip osteoarthritis

PurposeExacerbations of osteoarthritis (OA) pain significantly impair activities of daily living and walking, thereby reducing quality of life. Despite various treatments, persistent pain remains a challenge, particularly in patients with hip OA, where pain perception mechanisms extend beyond nociceptive pain. This study examined the pain characteristics in patients with hip OA, with an emphasis on neuropathic pain and central sensitization (CS). MethodsA total of 421 patients scheduled for unilateral total hip arthroplasty were analyzed using the painDETECT score and Central Sensitization Inventory to classify the pain types as nociceptive (Noc), CS, neuropathic (Neu), or neuropathic with CS (NeuCS). ResultsThe results indicated that 72.9% of patients experienced Noc pain, while 5.0% experienced combined NeuCS pain. The Neu and NeuCS groups exhibited significantly lower modified Harris Hip Score and EuroQol-5 Dimensions Questionnaire (EQ-5D) scores than the Noc group, indicating worse clinical outcomes and quality of life. Neu pain was more intense at rest, and when combined with CS it significantly diminished the quality of life. ConclusionThese findings highlight the need for mechanism-based pain treatment strategies, such as Neu pain and CS components, which independently and cumulatively affect clinical scores and quality of life. This study highlights the importance of early recognition and targeted treatment of these pain mechanisms in patients with hip OA. Future research should focus on evaluating therapeutic interventions for Neu pain and CS to optimize patient outcomes.

A neuronal marker of eye contact spontaneously activated in neurotypical subjects but not in autistic spectrum disorders

Attention to faces and eye contact are key behaviors for establishing social bonds in humans. In Autism Spectrum Disorders (ASD), a disturbance in neurodevelopment, impaired face processing and gaze avoidance are key clinical features for ASD diagnosis. The biological alterations underlying these impairments are not yet clearly established. Using high-density electroencephalography coupled with multi-variate pattern classification and group blind source separation methods we searched for face- and-face components-related neural signals that could best discriminate visual processing of neurotypical subjects (N = 38) from ASD participants (N = 27). We isolated a face-specific neural signal in the superior temporal sulcus peaking at 240 msec after face-stimulus onset. A machine learning algorithm applied on the extracted neural component reached 74% decoding accuracy at the same latencies, discriminating the neurotypical population from ASD subjects in whom this signal was weak. By manipulating attention on different parts of the face, we also found that the power of the evoked signal in neurotypical subjects varied depending on the region observed: it was strong when the eye region fell on the fovea to decrease on regions further away and outside the stimulus face. Such face and face-components selective neural modulations were not found in ASD, although they did show typical early face-related P100 and N170 signals. These results show that specialized cortical mechanisms for face perception show higher responses for eyes when attention is focused on gaze and that these mechanisms may be particularly affected in autism spectrum disorders.

Bilateral tDCS over the DLPFC enhances baroreceptor reflex sensitivity and inhibits blood pressure-related hypoalgesia

ObjectiveThis study investigated the impact of transcranial direct stimulation (tDCS) on pain perception, baroreflex sensitivity (BRS), and blood pressure (BP)-related hypoalgesia. MethodFifty-eight healthy participants were randomized to receive 1) bi-hemispheric tDCS over the dorsolateral prefrontal cortex (DLPFC) at 2 mA for 20 min, or 2) non-stimulation (Sham). Pain measures (threshold, tolerance, intensity and unpleasantness), emotional state (anxiety and mood), continuous BP, and electrocardiogram (ECG) data were recorded before, during, and after stimulation. ResultstDCS stimulation was followed by increases in BRS, pain intensity and unpleasantness. Anxiety decreased in the Sham group, but not in the tDCS group. Positive correlations between BP and pain threshold and tolerance before stimulation were observed. These remained during stimulation in the Sham group, but not in the tDCS group. Moreover, negative associations between BRS and BP only persisted in the Sham group. DiscussionThe results suggest that bilateral tDCS over the DLPFC enhances BRS and modulates pain perception and BP-related mechanisms. tDCS increases pain perception by inhibiting BP-related hypoalgesia and preventing habituation of anxiety. SignificanceLow BRS is a powerful prognostic factor of cardiovascular disease, such that its increase via tDCS may be a new therapeutic strategy for cardiovascular health promotion.

Microseconds-level coding for echo delay in the auditory brainstem of an FM-echolocating bat.

Echolocating big brown bats (Eptesicus fuscus) detect changes in ultrasonic echo delay with an acuity as sharp as 1 microsecond (μs) or less. How this perceptual feat is accomplished in the nervous system remains unresolved. Here, we examined the precision of latency registration (latency jitter) in neural population responses as a possible mechanism underlying the bat's hyperacuity. We recorded local field potentials in the cochlear nucleus and inferior colliculus of anesthetized big brown bats to sequences of sounds consisting of a simulated frequency modulated broadcast followed, at various echo delays, by a four-echo cascade. Latencies of the first negative response peak to the broadcast and to the first echo in the cascade were shorter in the cochlear nucleus than in the inferior colliculus, but latency jitter of this peak was comparable in both brainstem nuclei. Mean latency jitter, averaged over all stimulus conditions, was 51 μs in the cochlear nucleus and 56 μs in the inferior colliculus. Latency jitter to the successive echoes in the echo cascades was larger, with means of 125 μs and 111 μs, respectively. These values are lower than values commonly reported for single-neuron latency variability in bats and other mammals, and they approach within an order of magnitude the big brown bat's psychophysical performance. Latency jitter for synchronized population responses on a scale of microseconds reduces the gap between neurophysiological and behavioral measures of acuity. Further systems level analysis is necessary for understanding neural mechanisms of perception. XXXX XXXX.

Whole-brain responses to visual and auditory stimuli in anesthetized and minimally restrained awake mice using quiet zero echo time fMRI

Abstract Functional MRI (fMRI) is a flexible tool for sensory perception studies in animal models. However, animal fMRI studies are generally performed under anesthesia. Unfortunately, anesthesia affects brain function and sensory processing, complicating the interpretation of the findings. Since there is a growing need for fMRI protocols applicable for awake animals, we optimized a zero echo time Multi-Band Sweep Imaging with a Fourier Transformation (MB-SWIFT) fMRI approach for imaging awake mice. We implemented a 14-day habituation protocol that resulted in merely moderate motion of the mice while being head-fixed with the animals’ body and limbs being free-to-move. The sensory responsiveness between different states of consciousness was compared by imaging mice with visual and auditory stimulation schemes in the awake state and under ketamine-xylazine anesthesia. In awake mice, we observed a robust whole-brain activation of the ascending auditory and visual pathways, as well as higher sensory processing areas. Under ketamine-xylazine anesthesia, auditory responses were suppressed, and the temporal shapes of fMRI responses were different from those obtained in awake mice. Our results suggest that the quiet and motion-tolerant zero echo time MB-SWIFT approach allows complex behavioral fMRI designs in the awake state that promise to improve our understanding of the underlying mechanisms of perception.

Enhanced neural sensitivity to brief changes of happy over angry facial expressions in preschoolers: A fast periodic visual stimulation study.

Across childhood, emotion perception from facial expressions has traditionally been studied with event-related potentials (ERP). Here, we explored the novel fast periodic visual stimulation (FPVS) electroencephalography (EEG) approach to provide information about how brief changes in facial expressions are processed implicitly in young children's brains. Utilizing two FPVS tasks for the first time in preschoolers, we examined brain responses to (1) the discrimination of brief changes in facial expressions at maximum intensity and (2) thresholds for discrimination of gradual increasing facial expression intensities. Within a stream of neutral faces at 6 Hz, happy and angry faces were embedded with a frequency of 1.2 Hz. Additionally, children performed an emotion recognition task (ERT). Data were collected in the context of a training study for socio-emotional competencies with typically developing children (N = 74; 5.1[0.9] years; 34 females). FPVS data were collected post-training, where training was included as a controlling factor. Across FPVS tasks, we detected robust expression change responses, particularly with larger responses for happy versus angry faces in the maximum intensity task. ERT results paralleled neural findings with faster reaction times and higher accuracy rates for happy versus angry faces. For gradual increases in emotional intensity, we found linear increases in responses across emotions. The majority of the sample showed a significant expression change at 60% intensity. With its implicit nature, short duration, and robustness of individual responses, our results highlight the potential of FPVS in comparison to classical ERP methods to study neural mechanisms of emotion perception in preschool samples.

ILDIM-MFAM: interstitial lung disease identification model with multi-modal fusion attention mechanism.

This study aims to address the potential and challenges of multimodal medical information in the diagnosis of interstitial lung disease (ILD) by developing an ILD identification model (ILDIM) based on the multimodal fusion attention mechanism (MFAM) to improve the accuracy and reliability of ILD. Large-scale multimodal medical information data, including chest CT image slices, physiological indicator time series data, and patient history text information were collected. These data are professionally cleaned and normalized to ensure data quality and consistency. Convolutional Neural Network (CNN) is used to extract CT image features, Bidirectional Long Short-Term Memory Network (Bi-LSTM) model is used to learn temporal physiological metrics data under long-term dependency, and Self-Attention Mechanism is used to encode textual semantic information in patient's self-reporting and medical prescriptions. In addition, the multimodal perception mechanism uses a Transformer-based model to improve the diagnostic performance of ILD by learning the importance weights of each modality's data to optimally fuse the different modalities. Finally, the ablation test and comparison results show that the model performs well in terms of comprehensive performance. By combining multimodal data sources, the model not only improved the Precision, Recall and F1 score, but also significantly increased the AUC value. This suggests that the combined use of different modal information can provide a more comprehensive assessment of a patient's health status, thereby improving the diagnostic comprehensiveness and accuracy of ILD. This study also considered the computational complexity of the model, and the results show that ILDIM-MFAM has a relatively low number of model parameters and computational complexity, which is very favorable for practical deployment and operational efficiency.

Sub- and Supra-Second Duration Perception of Implied Motion: Differences Between Athletes and Non-Athletes.

This study aimed to investigate the differences in duration perception between athletes and non-athletes when looking at implied motion images within sub- and supra-second time ranges. By adopting the temporal bisection method, the study analyzed the duration perception of 20 college student athletes and 20 non-athletes regarding the implied motion of daily life (running and walking) and static postures (standing). The results showed that the effect of movement posture was significant, i.e., the perceived duration of the implied motion posture was longer than that of the static posture. Specifically, athletes perceived longer durations in the supra-second time range compared to non-athletes, indicating that long-term training enhanced athletes' time perception abilities. The findings provide new insights into the cognitive mechanisms of time perception and emphasize the influence of long-term physical training on temporal perceptual capabilities.

Possible Mechanism for Perception of Auditory-Verbal Hallucinations in Schizophrenia and Approaches to Their Weakening

We put forward a hypothesis that the processing and perception of auditory-verbal hallucinations in schizophrenia — internally generated speech, occurs in the same neural circuits as speech from external sources. These topographically organized cortico—basal ganglia—thalamocortical neural circuits include auditory, language, and frontal neocortical areas. It follows from our proposed mechanism for sound processing, that the increased action on dopamine D2 receptors on striatal neurons, which is considered the cause of hallucinations, should lead to a determined reorganization of activity in these neural circuits. As a result of this reorganization, the inhibition of neurons in the thalamic nuclei, including the internal geniculate body, should decrease synergistically through the direct and indirect pathways in the basal ganglia, and the excitation of connected with them cortical neurons, in the activity of which speech is represented, should increase. From this mechanism it follows that in order to weaken the perception of auditory-verbal hallucinations, it is necessary to increase the inhibition of thalamic neurons. Taking into account the known data on the distribution of receptors of different types on neurons in the striatum, thalamus and neocortex, as well as the previously formulated unified modification rules for the effectiveness of synaptic transmission in different structures, we proposed that agonists of adenosine A1 and muscarinic M4 receptors located on striatonigral spiny cells, giving rise to a direct disinhibitory pathway through the basal ganglia as well as antagonists of delta-opioid and cannabinoid CB1 receptors, located on striatopallidal spiny cells, giving rise to an indirect inhibitory pathway through the basal ganglia may be useful to weaken the perception of hallucinations. In addition, activation of A1 receptors can directly weaken the activity of neurons in the thalamus and neocortex due to the induction of long-term depression in the efficiency of their excitation. Inactivation of cannabinoid CB1 receptors on the projection GABAergic cells of the reticular thalamic nucleus may enhance their inhibitory effect on neurons of different thalamic nuclei. Since the proposed substances only indirectly affect the dopaminergic system, their use should not cause such pronounced side effects as D2 receptor antagonists, antipsychotics that are widely used to suppress auditory-verbal hallucinations.

Approximate planning in spatial search.

How people plan is an active area of research in cognitive science, neuroscience, and artificial intelligence. However, tasks traditionally used to study planning in the laboratory tend to be constrained to artificial environments, such as Chess and bandit problems. To date there is still no agreed-on model of how people plan in realistic contexts, such as navigation and search, where values intuitively derive from interactions between perception and cognition. To address this gap and move towards a more naturalistic study of planning, we present a novel spatial Maze Search Task (MST) where the costs and rewards are physically situated as distances and locations. We used this task in two behavioral experiments to evaluate and contrast multiple distinct computational models of planning, including optimal expected utility planning, several one-step heuristics inspired by studies of information search, and a family of planners that deviate from optimal planning, in which action values are estimated by the interactions between perception and cognition. We found that people's deviations from optimal expected utility are best explained by planners with a limited horizon, however our results do not exclude the possibility that in human planning action values may be also affected by cognitive mechanisms of numerosity and probability perception. This result makes a novel theoretical contribution in showing that limited planning horizon generalizes to spatial planning, and demonstrates the value of our multi-model approach for understanding cognition.

Better face-name recall is associated with better face recognition ability.

Face recognition is a highly developed and specialised human ability, distinct from other cognitive abilities. Previous studies examining individual differences in face recognition have focused on face perception and specialised perceptual mechanisms such as holistic face processing. However, the contribution of specific face memory processes to face recognition ability remains unclear. In 99 neurotypical individuals, we administered validated face perception assessments, three face memory tasks (old/new task, face-scene task, face-name/occupation task), and the Cambridge Face Memory Test (CFMT) to assess face recognition ability. We found that after accounting for face perception ability (which significantly predicted face recognition ability), face-name recall and recollection of faces in the face-scene task predicted unique variance in face recognition ability, with face-name recall being the strongest predictor. This highlights that associative memory mechanisms contribute to face recognition abilities and suggests that the ability to learn and recall proper names is particularly important to face recognition.

AI System for Autonomous Vehicles

The rapid advancement of artificial intelligence (AI) has revolutionized the development of autonomous vehicles, offering transformative potential for the future of transportation. This research investigates the implementation of AI-driven algorithms in autonomous vehicles, focusing on their ability to enhance decision-making, navigation, and safety. By employing state-of-the-art machine learning models, including deep learning and reinforcement learning, the study explores how these technologies can optimize real-time processing of sensor data, environmental perception, and adaptive control mechanisms. The findings demonstrate that AI algorithms can significantly improve the accuracy of object detection, trajectory prediction, and path planning, thereby reducing the likelihood of collisions and enhancing overall road safety. A key contribution of this work is the integration of a multi-modal sensor fusion approach, combining data from LiDAR, cameras, radar, and GPS to create a comprehensive and reliable understanding of the vehicle’s surroundings. Additionally, the research highlights the role of AI in enabling autonomous vehicles to learn from vast amounts of driving data, facilitating continuous improvement and adaptability in diverse driving conditions. The implications of this study are profound, suggesting that AI-powered autonomous vehicles could lead to safer, more efficient, and environmentally sustainable transportation systems. However, the research also identifies challenges related to computational complexity, real-time decision-making, and ethical considerations in AI-driven autonomy. Future work will focus on addressing these challenges and exploring the broader societal impacts of widespread autonomous vehicle adoption.

Unique cortical and subcortical activation patterns for different conspecific calls in marmosets.

The common marmoset (Callithrix jacchus) is known for its highly vocal nature, displaying a diverse range of calls. Functional imaging in marmosets has shown that the processing of conspecific calls activates a brain network that includes fronto-temporal areas. It is currently unknown whether different call types activate the same or different networks. In this study, nine adult marmosets (four females) were exposed to four common vocalizations (phee, chatter, trill, and twitter), and their brain responses were recorded using event-related fMRI at 9.4T. We found robust activations in the auditory cortices, encompassing core, belt, and parabelt regions, and in subcortical areas like the inferior colliculus, medial geniculate nucleus, and amygdala in response to these calls. Although a common network was engaged, distinct activity patterns were evident for different vocalizations that could be distinguished by a 3D convolution neural network, indicating unique neural processing for each vocalization. Our findings also indicate the involvement of the cerebellum and medial prefrontal cortex (mPFC) in distinguishing particular vocalizations from others.Significance Statement This study investigates the neural processing of vocal communications in the common marmoset (Callithrix jacchus). Utilizing event-related fMRI at 9.4T, we demonstrate that different calls (phee, chatter, trill, and twitter) elicit distinct brain activation patterns, challenging the notion of a uniform neural network for all vocalizations. Each call type distinctly engages various regions within the auditory cortices and subcortical areas. These findings offer insights into the evolutionary mechanisms of primate vocal perception and provide a foundation for understanding the origins of human speech and language processing.

Quantification of Visual Attention by Using Eye-Tracking Technology for Soundscape Assessment Through Physiological Response.

Because soundscapes affect human health and comfort, methodologies for evaluating them through physiological responses have attracted considerable attention. In this study, we proposed a novel method for evaluating visual attention by using eye-tracking technology to objectively assess soundscape perception. The study incorporated questionnaire surveys and physiological measurements focusing on visual attention responses. Results from the questionnaire indicated that perceptions of vehicles and the sky were 6% and 26% more sensitive, respectively, whereas perceptions of vegetation, based on physiological responses, were approximately 3% to 50% more sensitive. The soundscape quality prediction model indicates that the proposed methodology can complement conventional questionnaire-based models and provide a nuanced interpretation of eventfulness relationships. Additionally, the visual attention quantification technique enhanced the restoration responses of questionnaire-based methods by 1-2%. The results of this study are significant because the study introduces a novel methodology for quantifying visual attention, which can be used as a supplementary tool for physiological responses in soundscape research. The proposed method can uncover novel mechanisms of human perception of soundscapes that may not be captured by questionnaires, providing insights for future research in soundscape evaluation through physiological measurements.

Understanding cold stress response mechanisms in plants: an overview.

Low-temperature stress significantly impacts plant growth, development, yield, and geographical distribution. However, during the long-term process of evolution, plants have evolved complicated mechanisms to resist low-temperature stress. The cold tolerance trait is regulated by multiple pathways, such as the Ca2+ signaling cascade, mitogen-activated protein kinase (MAPK) cascade, inducer of CBF expression 1 (ICE1)-C-repeat binding factor (CBF)-cold-reulated gene (COR) transcriptional cascade, reactive oxygen species (ROS) homeostasis regulation, and plant hormone signaling. However, the specific responses of these pathways to cold stress and their interactions are not fully understood. This review summarizes the response mechanisms of plants to cold stress from four aspects, including cold signal perception and transduction, ICE1-CBF-COR transcription cascade regulation, ROS homeostasis regulation and plant hormone signal regulation. It also elucidates the mechanism of cold stress perception and Ca2+ signal transduction in plants, and proposes the important roles of transcription factors (TFs), post-translational modifications (PTMs), light signals, circadian clock factors, and interaction proteins in the ICE1-CBF-COR transcription cascade. Additionally, we analyze the importance of ROS homeostasis and plant hormone signaling pathways in plant cold stress response, and explore the cross interconnections among the ICE1-CBF-COR cascade, ROS homeostasis, and plant hormone signaling. This comprehensive review enhances our understanding of the mechanism of plant cold tolerance and provides a molecular basis for genetic strategies to improve plant cold tolerance.

Regulatory mechanisms of strigolactone perception in rice.

Strigolactones (SLs) are hormones essential for plant development and environmental responses. SL perception requires the formation of the complex composed of an SL receptor DWARF14 (D14), F-box protein D3, and transcriptional repressor D53, triggering ubiquitination and degradation of D53 to activate signal transduction. However, mechanisms of SL perception and their influence on plant architecture and environmental responses remain elusive and controversial. Here, we report that key residues at interfaces of the AtD14-D3-ASK1 complex are essential for the activation of SL perception, discover that overexpression of the D3-CTH motif negatively regulates SL perception to enhance tillering, and reveal the importance of phosphorylation and N-terminal disordered (NTD) domain in mediating ubiquitination and degradation of D14. Importantly, low nitrogen promotes phosphorylation and stabilization of D14 to repress rice tillering. These findings reveal a panorama of the activation, termination, and regulation of SL perception, which determines the plasticity of plant architecture in complex environments.

Ankle torque variance is a better indicator of balance control performance than plantar perceptual sensitivity threshold.

We explored whether ankle torque variability or plantar perceptual threshold explains human balance control more effectively. We hypothesized that ankle torque variance is a better indicator of center of pressure (COP) velocity variance than plantar perceptual sensitivity. Two conditions were tested: loaded (23-kg vest added) and unloaded, as loading should diminish plantar sensitivity and increase COP velocity variability. We created a linear feedback model to assess the noise change in the sensorimotor loop induced by loading. Plantar sensitivity was quantified using a psychophysical approach while participants stood barefoot. A linear motor applied a force impulse on the participant's heel. A "yes-no" method of limits was selected to identify plantar sole sensory thresholds in both conditions. We observed reduced plantar sensitivity in loaded compared with unloaded conditions. In the loaded condition, participants exhibited greater COP velocity variance, with significant positive Pearson's correlations confirming a substantial association between ankle torque and COP velocity variances for both loaded [variance accounted for (VAF): r2 = 44.56%, P = 0.018] and unloaded conditions (VAF: r2 = 58.83%, P = 0.004). No significant correlation existed between COP velocity variance and plantar sensitivity threshold for both loaded (VAF: r2 = 0.002%, P = 0.99) and unloaded conditions (VAF: r2 = 21.81%, P = 0.35). The model confirmed an ∼88% rise in sensorimotor loop noise in the loaded condition. Ankle torque variance assesses the precision of nonperceptual and perceptual detection mechanisms in evaluating whole body motions and the accuracy in converting sensory cues into ankle torque.NEW & NOTEWORTHY Plantar cutaneous information contributes to balance control by modulating motor commands, but plantar perceptual sensitivity is a suboptimal indicator of balance performance. Multiple sensory cues encode whole body dynamics, guiding sensorimotor mechanisms to minimize body sway variability. Ankle torque variance is proposed as a superior measure for explaining balance control performance and evaluating the sensorimotor loop's functioning in balance control.

Enhancing small target traffic sign detection with ML_SAP in YOLOv5s

Detecting small traffic signs poses significant challenges due to the complex nature and dynamic conditions of real-world traffic scenarios. In response to these challenges, we propose an improved YOLOv5s target detection model incorporating the multilevel squeeze feature perception (ML_SAP) mechanism, aiming to increase the accuracy of small traffic sign detection. First, an additional detection layer is incorporated to enhance the model’s capacity for detecting small-scale traffic signs. This improvement is accompanied by the adoption of a WIoU loss function, which evaluate the quality of anchor boxes. Moreover, the ML_SAP mechanism is designed to promotes the fusion and extraction of features at different levels. This mechanism effectively increases the network model’s proficiency in identifying small targets under varying environmental conditions. To verify the effectiveness of the improved method, we conducted extensive experiments on two public transportation sign datasets. Notably, on the challenging samples in the CCTSDB-2021 dataset, the improved model achieves a detection recall of 77.1%, which is 5.4% higher than that of YOLOv5s, and a mean average precision (mAP) of 82.7%, which is 3.9% higher than that of the base model. Furthermore, the model achieves a detection recall of 91.3% on the TT100K dataset, which is 3.7% higher than the performance of YOLOv5s, and a mean precision (mAP) of 91.5%, which is 4.6% higher than that of the base model.