Keren He

Dr. Keren He


Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 August 2026 in Information
Verified
Journal Article
Open access
PDF file
YGYufei Gao
Yufei Gao
QWQian Wu
Qian Wu
SZShuliang ZHU
Shuliang ZHU
KHKeren He
Keren He
WWWei Weng
Wei Weng
JZJinjia Zhou
Jinjia Zhou
Music-to-dance generation requires precisely aligning movement dynamics with musical rhythm, yet existing methods rely on shallow conditioning or auxiliary beat-alignment objectives that fail to establish stable beat–motion correspondences. We present ChoreDiffusion, a diffusion-based framework that integrates explicit beat guidance directly into the denoising process. Central to our approach is a beat-enhanced cross-modal attention mechanism that injects beat-salience cues at every refinement step, promoting fine-grained synchronization beyond the reach of conventional conditioning pipelines. To support multiple dance styles within a unified model, we incorporate lightweight low-rank adaptation (LoRA) modules that encode style-specific motion signatures with only a small set of additional parameters per style, and a three-stage progressive curriculum stabilizes the joint learning of rhythmic alignment and stylistic expressivity. Experiments on two public multi-style dance benchmarks (AIST++ and FineDance) show that ChoreDiffusion achieves the lowest FID values among the compared generation methods on both benchmarks, while maintaining competitive rhythm alignment and multi-style controllability. These results indicate that embedding beat-aware guidance during generation, rather than applying it afterwards, is an effective route toward human-like musicality in music-driven choreography.

1 June 2026 in Sensors
Verified
Journal Article
Open access
PDF file
KHKeren He
Keren He
KXKun XIANG
Kun XIANG
YGYufei Gao
Yufei Gao
YYYang Yu
Yang Yu
JZJinjia Zhou
Jinjia Zhou
With the increasing demand for video processing in both human perception and machine vision applications, enhancing heavily compressed video has become a critical problem in practical multimedia systems. In many real-world scenarios, video data acquired by image sensors are often compressed for efficient transmission and storage, which introduces compression artifacts and degrades both visual quality and downstream task performance. This issue is especially significant in sensor-based systems such as surveillance cameras and mobile imaging devices. To address these challenges, we propose a novel joint human–machine video enhancement framework for compressed video enhancement that jointly targets human perceptual quality and machine vision performance. The framework integrates four complementary components: a Spatio-Temporal Fusion Module that leverages inter-frame correlations, a High-Frequency Semantic Fusion module for recovering structurally important details relevant to machine tasks, a Texture-Guided Model that enhances low-level visual features, and a Refined Attention Residual Quality Enhancement Module that adaptively emphasizes salient regions. By progressively combining these modules, the framework effectively restores compressed content while preserving task-relevant semantics. The experimental results demonstrate that our method consistently outperforms existing approaches, achieving higher PSNR and SSIM as well as improved object detection and video object segmentation performance. These results highlight the framework’s practical applicability for compressed video enhancement in sensor-based systems, including intelligent surveillance and autonomous imaging platforms.

28 February 2026 in Sensors
Verified
Journal Article
Open access
PDF file
KHKeren He
Keren He
YGYufei Gao
Yufei Gao
QWQi Wang
Qi Wang
HWHaixin Wang
Haixin Wang
JZJinjia Zhou
Jinjia Zhou
Down-sampling-based video compression frameworks have shown great potential in improving compression efficiency in modern sensing and imaging systems. However, existing methods ignore critical spatial and temporal redundancy, and treat all frames uniformly during down-sampling. This leads to the loss of important information and impacts compression efficiency. To address these limitations, this paper proposes a temporal down-sampling system, in which only intermediate frames are down-sampled while preserving key frames with high quality for reference. On the decoding side, we employ a frame-recurrent enhancement mechanism to maximize the use of temporal redundancy information. In the fusion of enhancement stage, we design a Multi-scale Temporal-Spatial Attention (MTSA) module. MTSA consists of two components: Multi-Temporal Attention (MTA) and Pyramid Spatial Attention (PSA). MTA performs multi-scale temporal correlation modeling, expanding the receptive field and providing stable cues in compressed regions. PSA integrates local spatial saliency and contextual structure in a progressive and multi-stage manner. Extensive experiments show that our approach achieves consistent BD-rate reductions. Under All-Intra, Low-Delay-P, and Random Access configurations, we observe BD-rate reductions of I, P, and B frames ranging from 14% to 39% compared to VVC, and outperform prior approaches anchored by the standard HEVC.

21 April 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
QWQi Wang
Qi Wang
KHKeren He
Keren He
SZShuliang ZHU
Shuliang ZHU
JZJinjia Zhou
Jinjia Zhou
Motivated by the limitation that most contrastive knowledge distillation methods optimize only one perspective, this paper proposes a simple and efficient contrastive knowledge distillation method that formulates distillation as a dual-perspective objective over the class-wise and the sample-wise. Unlike prior contrastive knowledge distillation, we jointly align teacher and student logit distributions at both perspectives. Concretely, same input pairs serve as anchors whose similarities are optimized against negatives drawn from other classes, enforcing class-wise alignment and maximizing separability between classes. To counterbalance the potential distortion of sample geometry caused by pure class-wise alignment, we simultaneously perform sample-wise alignment, matching each student logit only to its teacher counterpart for the same input, irrespective of class relations. This dual alignment supplies complementary supervision, stabilizes sample-wise structure, and sharpens decision boundaries. Moreover, we add a label-guided same class contrast that promotes compactness within classes while further enhancing separability between classes. We further propose a delayed learning strategy that orchestrates multiple contrastive losses, enabling a more effective integration that preserves their respective advantages. Across CIFAR-100, ImageNet, Tiny-ImageNet, and STL-10, our approach surpasses most logit-based and feature-based baselines in image classification, transfer learning, and ViT distillation, while maintaining superior training efficiency.