We propose Cross-Attention in Audio, Space, and Time (C$\text{A}^{2}$A2ST), a transformer-based method for holistic video recognition. Recognizing actions in videos requires both spatial and temporal understanding, yet most existing models lack a balanced spatio-temporal understanding of videos. To address this, we propose a novel two-stream architecture, called Cross-Attention in Space and Time (CAST), using only RGB input. In each layer of CAST, Bottleneck Cross-Attention (B-CA) enables spatial and temporal experts to exchange information and make synergistic predictions. For holistic video understanding, we extend CAST by integrating an audio expert, forming Cross-Attention in Visual and Audio (CAVA). We validate the CAST on benchmarks with different characteristics, EPIC-KITCHENS-100, Something-Something-V2, Kinetics-400, ActivityNet, and HD-EPIC to show balanced performance. We also validate the CAVA on audio-visual action recognition benchmarks, including UCF-101, VGG-Sound, KineticsSound, EPIC- SOUNDS, and HD-EPIC-SOUNDS. CAVA shows favorable performance on these datasets, demonstrating the effective information exchange among multiple experts within the B-CA module. In addition, C$\text{A}^{2}$A2ST combines CAST and CAVA by employing spatial, temporal, and audio experts through cross-attention, achieving balanced and holistic video understanding.

Effective explanations of video action recognition models should disentangle how movements unfold over time from the surrounding spatial context. However, existing methods—based on saliency—produce entangled explanations, making it unclear whether predictions rely on motion or spatial context. Language-based approaches offer structure but often fail to explain motions due to their tacit nature—intuitively understood but difficult to verbalize. To address these challenges, we propose Disentangled Action aNd Context concept-based Explainable (DANCE) video action recognition, a framework that predicts actions through disentangled concept types: motion dynamics, objects, and scenes. We define motion dynamics concepts as human pose sequences. We employ a large language model to automatically extract object and scene concepts. Built on an ante-hoc concept bottleneck design, DANCE enforces prediction through these concepts. Experiments on four datasets—KTH, Penn Action, HAA500, and UCF101—demonstrate that DANCE significantly improves explanation clarity with competitive performance. Through a user study, we validate the superior interpretability of DANCE. Experimental results also show that DANCE is beneficial for model debugging, editing, and failure analysis.

Recognizing human actions in videos requires spatial and temporal understanding. Most existing action recognition models lack a balanced spatio-temporal understanding of videos. In this work, we propose a novel two-stream architecture, called Cross-Attention in Space and Time (CAST), that achieves a balanced spatio-temporal understanding of videos using only RGB input. Our proposed bottleneck cross-attention mechanism enables the spatial and temporal expert models to exchange information and make synergistic predictions, leading to improved performance. We validate the proposed method with extensive experiments on public benchmarks with different characteristics: EPIC-Kitchens-100, Something-Something-V2, and Kinetics-400. Our method consistently shows favorable performance across these datasets, while the performance of existing methods fluctuates depending on the dataset characteristics. The code is available at https://github.com/KHU-VLL/CAST.

Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) is a useful tool for rapid identification of microorganisms based on the protein mass profile represented in a mass spectrum of the microorganism. Typically, markers that are specific for particular microorganisms are extracted from the mass information obtained by MALDI-TOF MS, and a machine learning technique is applied to the markers. Identification of mycobacteria is of high clinical importance in that different pathogens must be treated with different antibiotics, but is still challenging because spectral patterns of different mycobacteria appear similar. In this paper, we propose a novel approach to use both positive and negative markers in order to enhance discrimination between the spectral patterns of different mycobacteria. We apply the proposed method to classify species in the Mycobacterium abscessus and Mycobacterium fortuitum groups. Experimental results demonstrate that, when combined with various classifier techniques, our method significantly improves the accuracy of mycobacteria identification.

MALDI-TOF mass spectrometry has high social and economic value in rapid identification of microorganisms based on the protein mass profile represented in a mass spectrum of the microorganism. Numerous studies have been conducted to identify microorganisms using MALDI-TOF MS. Markers are characteristics that can be used to uniquely distinguish microorganisms. Microorganisms can be identified by applying markers selected based on the extracted mass information. Previous studies demonstrated that combining mass information extracted by MALDI-TOF MS with machine-learning techniques can improve microorganism classification. Classification of microorganisms is particularly difficult and critical for mycobacteria because various pathogens should be treated with different prescriptions, although they exhibit similar compositions. It is quite challenging to accurately identify mycobacteria using conventional methods because their MALDI-TOF MS patterns are similar to each other. In this study, we propose a support vector machine model for improving the distinction of similar species by learning positive and negative markers separately extracted in each group. We classified species in the Mycobacterium abscessus group and Mycobacterium fortuitum group. Our novel approach applies negative markers to classify similar species and improves the identification of similar species using a combination of positive and negative markers.

In the Internet of Things (IoT) era, people are very interested in wearable devices such as smart watches. These devices measure individual physiological time series such as blood pressure, heart rate, and EEG. With this functionality, people can check the status of their own health. This healthcare service usually sends individual physiological time series to remote clusters for calculation. A remote healthcare service is particularly necessary for patients suffering from chronic and urgent diseases such as cardiovascular disease. It is also necessary to predict urgent signals for proper treatment. One method to predict urgent signals is by clustering physiological time series and comparing the new physiological time series with the previous time series in a cluster. It means searching the time series similar to risk features. In other words, the detection and comparison of features in time series are important. Therefore, in this study, we propose a biosignal processing system based on the Haar transform of time series in IoT applications. We discuss the validity of this system according to various perspectives. The Haar transform of a time series reflects the trend of the time series; thus, we can recognize the trend of the time series more easily. In addition, we can reduce the storage size of the time series. This is especially helpful because the volume of a time series is massive in the IoT era. Although the reduction of information in a time series can distort the similarity accuracy, it does not distort it significantly.