This paper proposes a Context-adaptive Audio-Visual Neural Network (CAVN) model for anomaly detection in public safety systems. Existing approaches primarily rely on visual data and employ simple fusion strategies for combining modalities, which leads to limitations in capturing complex semantic relationships. The proposed model consists of four main components: a visual feature extraction module based on SlowFast architecture, an audio feature extraction module based on Audio Spectrogram Transformer (AST), a fusion module based on bidirectional cross-attention mechanism, and a temporal context aggregation module based on Transformer encoder. The main scientific novelty of the model lies in the adaptive modality balancing mechanism, which dynamically adjusts the relative importance of modalities under different conditions (dark/bright, noisy/quiet). Experimental results demonstrate that the proposed CAVN model outperforms existing methods by in overall accuracy and by in dark conditions. Ablation studies confirmed the contribution of each module to the overall performance.
| Mualliflar | Ravshanov, N., Boborakhimov, B.I., Berdiev, M.I., Равшанов, Н., Боборахимов, Б.И., Бердиев, М.И. |
|---|---|
| Jurnal | Ҳисоблаш ва амалий математика муаммолари |
| Nashr sanasi | 2026-01-11 |
| Son | 6 |
| Betlar | 88-102 |
| Til | Ingliz |
| DOI | 10.71310/pcam.6_70.2025.07 |
DOI: 10.71310/pcam.6_70.2025.07 · Maqolaning asl sahifasi
dynamic weighting, spatiotemporal representation learning, attention-based alignment, robust anomaly recognition, real-world surveillance, динамическое взвешивание модальностей, пространственно-временное представление, выравнивание на основе внимания, устойчивое распознавание аномалий, видеонаблюдение в реальных условиях
For the first time, a fractal solar collector operating in the cogeneration mode is proposed. In this context, we are talking about a modern innovative system that uses fractal structures to optimize the processes of…
In this paper, the Cauchy problem with boundary conditions for an inhomogeneous parabolic equation with changing time direction is investigated. To construct the solution, the method of separation of variables (Fourier…
The paper addresses a numerical solution of a plane–radial boundary inverse problem for the relaxation filtration equation describing fluid flow in an elastically deformable porous medium. The study is motivated by the…
The paper models the spatiotemporal dynamics of the Kattakurgan Reservoir area using satellite-based water and vegetation indices combined with ensemble machinelearning methods. Multi-year time series of NDWI, NDVI, and…
Hypersingular integral equations are encountered in a number of fields, such as the dynamics of air and fluid flow, elasticity, and wave propagation theory. Analytical solutions for these integral equations exist, but…
This article proposes a dynamic model of a coupled thermoelasticity problem under stress. A coupled boundary value problem is formulated, consisting of three differen tial equations for the stress and temperature tensor…
Integrals of rapidly oscillating functions arise mainly in the theory of special functions and Fourier analysis, but also in other applied and computational sciences and engineering, for example, in theoretical physics…
The article analyzes the hydrodynamic processes associated with underground mining, in particular, the acid extraction of precious metals from ore deposits. In order to comprehensively study, monitor and predict the…
For the numerical solution of the Dirichlet problem for the Poisson equation, both direct and iterative methods have been developed. However, the required number of arithmetic operations for direct methods, as well as…
This paper presents the design and analysis of an innovative solar-powered water supply system for multi-storey residential buildings under conditions of high solar insolation and limited energy resources. Instead of…