Multimodal Sentiment Analysis(MSA)seeks to predict a speaker's sentiment orientation by comprehensively utilizing modalities such as text,vision,and audio.As deep learning and cross-modal fusion technologies evolv...Multimodal Sentiment Analysis(MSA)seeks to predict a speaker's sentiment orientation by comprehensively utilizing modalities such as text,vision,and audio.As deep learning and cross-modal fusion technologies evolve,key challenges include alleviating heterogeneity across modality feature spaces,avoiding bias from fixed main-modal fusion strategies,and enhancing model adaptability to dynamic changes in modality contribution across different samples.To address these issues,this paper proposes a multimodal sentiment analysis framework based on adaptive modality selection and contrastive learning alignment,named Adaptive Modality Selection and Guided Fusion Network(AMSGFN).The framework first employs a crossmodal contrastive learning alignment mechanism to map text,vision,and audio features into a shared semantic space,mitigating semantic discrepancies among heterogeneous modalities.A lightweight modality scoring module then evaluates the discriminability and reliability of each modality for the current sample,adaptively identifying the dominant modality.Building on this,a dominant modality-guided fusion mechanism selectively integrates supplementary information from auxiliary modalities around the dominant modality,highlighting key emotional semantics while suppressing noise and redundant information.Experimental results demonstrate that the proposed method achieves superior performance compared to existing approaches across multiple public datasets,confirming the effectiveness and robustness of the framework in multimodal sentiment analysis.展开更多
Mode-localized sensors have attracted significant attention due to their exceptional sensitivity and inherent ability to reject common-mode noise.This high sensitivity arises from the substantial shifts in resonator a...Mode-localized sensors have attracted significant attention due to their exceptional sensitivity and inherent ability to reject common-mode noise.This high sensitivity arises from the substantial shifts in resonator amplitudes induced by energy confinement in weakly coupled resonators.Despite their promising attributes,there has been limited research on the mechanisms of energy confinement.This paper presents both qualitative and quantitative analyses of energy confinement within weakly coupled resonators and concludes them as the concept of modal dominance.This concept elucidates that mode frequencies are predominantly dictated by the natural frequencies of the internal resonators,facilitating spatial energy confinement.Based on this modal dominance,a novel concept of virtually coupled resonators is proposed,which obviates the need for physical coupling structures.Instead,energy confinement is achieved through a frequency offset between two independent resonators,resulting in a similar amplitude ratio output and enhanced sensitivity.To further enhance performance,a double-closed-loop control scheme is developed for virtually coupled resonators,expanding the bandwidth in comparison to weakly coupled resonators.Experimental results validate the feasibility of virtually coupled resonators and the double-closed-loop control,demonstrating a 2.7-fold improvement in amplitude ratio sensitivity and at least a four-fold enhancement in bandwidth relative to weakly coupled resonators with identical parameters.展开更多
文摘Multimodal Sentiment Analysis(MSA)seeks to predict a speaker's sentiment orientation by comprehensively utilizing modalities such as text,vision,and audio.As deep learning and cross-modal fusion technologies evolve,key challenges include alleviating heterogeneity across modality feature spaces,avoiding bias from fixed main-modal fusion strategies,and enhancing model adaptability to dynamic changes in modality contribution across different samples.To address these issues,this paper proposes a multimodal sentiment analysis framework based on adaptive modality selection and contrastive learning alignment,named Adaptive Modality Selection and Guided Fusion Network(AMSGFN).The framework first employs a crossmodal contrastive learning alignment mechanism to map text,vision,and audio features into a shared semantic space,mitigating semantic discrepancies among heterogeneous modalities.A lightweight modality scoring module then evaluates the discriminability and reliability of each modality for the current sample,adaptively identifying the dominant modality.Building on this,a dominant modality-guided fusion mechanism selectively integrates supplementary information from auxiliary modalities around the dominant modality,highlighting key emotional semantics while suppressing noise and redundant information.Experimental results demonstrate that the proposed method achieves superior performance compared to existing approaches across multiple public datasets,confirming the effectiveness and robustness of the framework in multimodal sentiment analysis.
基金supported by the National Science Foundation of China(No.52435012 and No.52475606)the National Key Research and Development Program of China(No.2023YFB3208800)+2 种基金Innovation Capability Support Program of Shaanxi(No.2024RS-CXTD-7)the Key Research and Development Program of Shaanxi Province(2024GX-YBXM-193)the Fundamental Research Funds for the Central Universities.
文摘Mode-localized sensors have attracted significant attention due to their exceptional sensitivity and inherent ability to reject common-mode noise.This high sensitivity arises from the substantial shifts in resonator amplitudes induced by energy confinement in weakly coupled resonators.Despite their promising attributes,there has been limited research on the mechanisms of energy confinement.This paper presents both qualitative and quantitative analyses of energy confinement within weakly coupled resonators and concludes them as the concept of modal dominance.This concept elucidates that mode frequencies are predominantly dictated by the natural frequencies of the internal resonators,facilitating spatial energy confinement.Based on this modal dominance,a novel concept of virtually coupled resonators is proposed,which obviates the need for physical coupling structures.Instead,energy confinement is achieved through a frequency offset between two independent resonators,resulting in a similar amplitude ratio output and enhanced sensitivity.To further enhance performance,a double-closed-loop control scheme is developed for virtually coupled resonators,expanding the bandwidth in comparison to weakly coupled resonators.Experimental results validate the feasibility of virtually coupled resonators and the double-closed-loop control,demonstrating a 2.7-fold improvement in amplitude ratio sensitivity and at least a four-fold enhancement in bandwidth relative to weakly coupled resonators with identical parameters.
