Frame acquisition is crucial to the entire receiving process. However, both traditional methods and existing deep learning approaches encounter performance challenges in low signal-to-noise ratio (SNR) scenarios. To address this issue, this letter proposes a novel deep learning-enabled frame acquisition scheme. We first introduce a data augmentation strategy that incorporates random time offsets and additive noise, specifically designed to mimic non-ideal reception conditions, such as capturing an incomplete preamble. Subsequently, we propose a frame acquisition neural network based on an efficient large kernel module, namely ELKFANet. This module decouples a standard large convolutional kernel into a sequence of horizontal, vertical, and pointwise convolutions, thereby capturing periodic and structural signal features through a global receptive field while maintaining computational efficiency. Experimental results show that under low SNR conditions ranging from −20 dB to 0 dB, ELKFANet achieves superior overall detection performance compared with existing methods across different channel models. Furthermore, the proposed module strikes an effective balance between computational complexity and detection performance, and exhibits strong scalability across different bandwidth scenarios.
With the rapid advancement of generative artificial intelligence, deepfake speech has emerged as a significant threat to digital audio authenticity, posing challenges for forensic analysis and legal applications. In this study, we propose EMBNet, a task-oriented deepfake speech detection framework that integrates efficient channel attention (ECA) with a multi-scale bottleneck to enhance the representation of subtle acoustic anomalies. The ECA module adaptively emphasizes critical feature channels, while the multi-scale bottleneck captures local and hierarchical spoofing traces across multiple time-frequency resolutions. The proposed framework effectively balances fine-grained local detail modeling with global hierarchical representation, improving the detection of weakly manifested spoofing patterns. Extensive experiments on the ASVspoof 2019 Logical Access dataset demonstrate that EMBNet significantly outperforms existing baseline and state-of-the-art methods, achieving an EER of 2.67%, an AUC of 97.32%, and an F1-score of 97.28%. Ablation studies further confirm the complementary contributions of the ECA and multi-scale modules to overall performance. The proposed approach demonstrates promising performance for forensic audio analysis and may support forensic audio authenticity assessment by improving the discrimination between genuine and manipulated speech.
Haitao Yang, Fen Li, Xin Cai et al.· Journal of Forensic Sciences· 0 citations
Deep joint source-channel coding (DeepJSCC) has emerged as a promising paradigm for semantic-aware wireless image transmission, achieving strong performance under challenging channel conditions. However, MSE-trained DeepJSCC systems typically achieve high peak signal-to-noise ratio (PSNR) values but suppress high-frequency details, resulting in perceptually blurry reconstructions that fail to capture fine textures and edge information. Existing perceptual enhancement approaches for JSCC systems face significant practical limitations: full transceiver redesign methods require replacing both the transmitter and the receiver with large models (19–31 million parameters), incurring substantial deployment costs; diffusion-based refinement approaches require over 1700 million additional parameters and introduce inference latency exceeding 13 s, rendering them unsuitable for latency-constrained wireless applications; and generic image restoration networks lack channel state awareness and cannot adapt to varying signal-to-noise ratio (SNR) conditions. This paper proposes a lightweight receiver-only perceptual enhancer designed for use with frozen DeepJSCC backbones. The proposed module adopts residual learning with feature-wise linear modulation (FiLM)-based SNR-adaptive modulation to dynamically adjust the enhancement strength under varying channel conditions. A radially weighted FFT magnitude loss is further introduced to guide high-frequency recovery. The enhancer adds only 0.29 million trainable parameters (<1% of the backbone) and requires neither transmitter modification nor backbone retraining. Extensive experiments on the Kodak24 and DIV2K datasets demonstrate a 34.4–37.5% LPIPS reduction over the frozen DeepJSCC baseline under AWGN channels. Supplementary robustness evaluations further show a 30–33% LPIPS reduction under Rayleigh fading, and stable generalization to unseen SNR levels. The receiver-side decoder-plus-enhancer pipeline requires 43 ms at 768 × 512 resolution, corresponding to approximately 23 frames per second.
Shouquan Hou, Peng Zhao, Nuo Chen· Italian National Conference...· 0 citations
The rapid advancement of voice synthesis technologies such as text-to-speech and voice conversion poses significant threats to speech-based authentication systems, necessitating robust deepfake detection methods. In this work, we propose a novel E-Branchformer-based architecture that effectively leverages self-supervised speech representations for audio deepfake detection. Our model employs parallel branches to simultaneously capture global contextual dependencies through multi-head self-attention and local temporal patterns through convolutional processing. To enhance discriminative capability, we integrate depthwise convolution and Squeeze-and-Excitation modules that enrich the classification token with refined patch token information after feature merging. Extensive experiments on ASVspoof 2021 LA, DF, and In-the-Wild datasets demonstrate state-of-the-art performance with equal error rates of 0.88%, 1.85%, and 6.30% respectively, substantially outperforming existing methods. Comprehensive ablation studies validate that the dual-branch architecture provides complementary discriminative information, Squeeze-and-Excitation Aggregation significantly improves SSL feature integration, and the combination of DWConv and SE modules is critical for effective class token enhancement. The superior performance on real-world scenarios demonstrates strong generalization capability to diverse acoustic conditions and unseen spoofing attacks.
Phuong Dat, Học Thủ, T. Nguyễn et al.· 0 citations
DSF-Net is a novel neural network that introduces a dual-strategy fusion approach for the AVSELD task, designed for robust and computationally efficient multi-modal comprehension.
This paper proposes BinRVR, a binarized RAW video restoration framework that reduces computation and parameters by approximately 96% while incurring only about 4% performance degradation, and develops a Distribution-Aware Binarized Convolution (DAB-Conv) that leverages the statistics of full-precision activations to mitigate quantization errors.
Tianyu Zhu, Ying Fu, Hesong Li et al.· IEEE Transactions on Pattern...· 0 citations
Infrared Small Target Detection (ISTD) is crucial for applications such as traffic monitoring and maritime surveillance. However, it remains highly challenging due to weak target signals and the absence of rich texture information, often resulting in low detection accuracy. Existing deep learning-based ISTD methods typically struggle to balance the trade-off between modeling long-range dependencies and avoiding feature oversmoothing. To this end, we propose RSFNet, a retention-based network with spatial–frequency joint enhancement for ISTD. RSFNet introduces a bidirectional 2D decay–retention attention mechanism into the vision Transformer (ViT) framework, which effectively suppresses background noise while capturing long-range dependencies. In addition, we design a Spatial–Frequency Joint Enhancement Module (SFE) to facilitate the transfer of salient target features from the encoder to the decoder. SFE integrates spatial and frequency domain features to facilitate global–local information interaction. Extensive experiments conducted on multiple publicly available ISTD datasets demonstrate that RSFNet significantly outperforms state-of-the-art (SOTA) methods in both detection accuracy and training efficiency, with a nearly 32% reduction in training time.
