A5005515553- Music and Audio Processing
- Biomedical Text Mining and Ontologies
- Emotion and Mood Recognition
- Video Analysis and Summarization
- Topic Modeling
- Obstructive Sleep Apnea Research
- Non-Invasive Vital Sign Monitoring
- Speech and Audio Processing
- Advanced MIMO Systems Optimization
- Sleep and Wakefulness Research
- Wikis in Education and Collaboration
- Energy Harvesting in Wireless Networks
- Advanced Image and Video Retrieval Techniques
- Digital Media Forensic Detection
- Wireless Power Transfer Systems
- Wireless Body Area Networks
- Text Readability and Simplification
- Advanced Wireless Communication Technologies
- Natural Language Processing Techniques
- Context-Aware Activity Recognition Systems
- Indoor and Outdoor Localization Technologies
- IoT Networks and Protocols
- Human Pose and Action Recognition
- Speech Recognition and Synthesis
- Multimodal Machine Learning Applications
Roche (Switzerland)
2023
Koç University
2018-2022
This study presents the outcomes of shared task competition BioCreative VII (Task 3) focusing on extraction medication names from a Twitter user's publicly available tweets (the 'timeline'). In general, detecting health-related is notoriously challenging for natural language processing tools. The main challenge, aside informality used, that people tweet about any and all topics, most their are not related to health. Thus, finding those in timeline mention specific concepts such as...
Radio Frequency Energy Harvesting (RF-EH) networks are key enablers of massive Internet-of-things by providing controllable and long-distance energy transfer to energy-limited devices. Relays, helping either or information transfer, have been demonstrated significantly improve the performance these networks. This paper studies joint relay selection, scheduling, power control problem in multiple-source-multiple-relay RF-EH under nonlinear EH conditions. We first obtain optimal solution...
The increasing volume of user-generated human-centric video content and its applications, such as retrieval browsing, require compact representations addressed by the summarization literature. Current supervised studies formulate a sequence-to-sequence learning problem, existing solutions often neglect surge view, which inherently contains affective content. In this study, we investigate affective-information enriched task for videos. First, train visual input-driven state-of-the-art...
Emerging technologies enforce strict requirements on future wireless networks such as massive connectivity that cannot be supported with scheduled access. Contention based Non-Orthogonal Multiple Access is a novel technique to overcome by efficient use of resources. However, most the solutions proposed in this direction assumes different loads which would degrade performance significantly if they not hold. To stress these assumptions resource efficiency metric defined and state art are...
In this study, we focus on continuous emotion recognition using body motion and speech signals to estimate Activation, Valence, Dominance (AVD) attributes. Semi-End-To-End network architecture is proposed where both extracted features raw are fed, trained multi-task learning (MTL) rather than the state-of-the-art single task (STL). Furthermore, correlation losses, Concordance Correlation Coefficient (CCC) Pearson (PCC), used as an optimization objective during training. Experiments conducted...
Video summarization attracts attention for efficient video representation, retrieval, and browsing to ease volume traffic surge problems. Although mostly uses the visual channel compaction, benefits of audio-visual modeling appeared in recent literature. The information coming from audio can be a result correlation content. In this study, we propose new framework integrating four ways fusion with GRU-based attention-based networks. Furthermore, investigate explainability methodology using...
Minimum length scheduling is used to ensure the strict delay requirements of time-critical applications in wireless powered communications networks (WPCNs). The previous optimal and sub-optimal solutions problem suffer from run-time complexity iterative algorithms, which makes real-time unpractical. This paper proposes a deep learning based framework for low-complexity solution minimum half-duplex WPCNs. objective minimize duration schedule energy harvesting (EH) information transmission...
Video summarization attracts attention for efficient video representation, retrieval, and browsing to ease volume traffic surge problems. Although mostly uses the visual channel compaction, benefits of audio-visual modeling appeared in recent literature. The information coming from audio can be a result correlation content. In this study, we propose new framework integrating four ways fusion with GRU-based attention-based networks. Furthermore, investigate explainability methodology using...
The BioCreative VII Track 3 challenge focused on the identification of medication names in Twitter user timelines. For our submission to this challenge, we expanded available training data by using several augmentation techniques. augmented was then used fine-tune an ensemble language models that had been pre-trained general-domain content. proposed approach outperformed prior state-of-the-art algorithm Kusuri and ranked high competition for selected objective function, overlapping F1 score.
As speech interfaces are getting richer and widespread, emotion recognition promises more attractive applications. In the continuous (CER) problem, tracking changes across affective states is an essential desired capability. Although CER studies widely use correlation metrics in evaluations, these do not always capture all high-intensity domain. this paper, we define a novel burst detection problem to of attributes accurately. We formulate two-class classification approach isolate regions...
Sleep is restoration process of the body. The efficiency this directly correlated to amount time spent at each sleep phase. Hence, automatic tracking via wearable devices has attracted both researchers and industry. Current state-of-the-art solutions are memory processing greedy they require cloud or mobile phone connectivity. We propose a efficient architecture which can work in embedded environment without needing any connection. In study, novel proposed that consists feature extraction...
Sleep is restoration process of the body. The efficiency this directly correlated to amount time spent at each sleep phase. Hence, automatic tracking via wearable devices has attracted both researchers and industry. Current state-of-the-art solutions are memory processing greedy they require cloud or mobile phone connectivity. We propose a efficient architecture which can work in embedded environment without needing any connection. In study, novel proposed that consists feature extraction...
Increasing volume of user-generated human-centric video content and their applications, such as retrieval browsing, require compact representations that are addressed by the summarization literature. Current supervised studies formulate a sequence-to-sequence learning problem existing solutions often neglect surge view, which inherently contains affective content. In this study, we investigate affective-information enriched task for videos. First, train visual input-driven state-of-the-art...
As speech-interfaces are getting richer and widespread, speech emotion recognition promises more attractive applications. In the continuous (CER) problem, tracking changes across affective states is an important desired capability. Although CER studies widely use correlation metrics in evaluations, these do not always capture all high-intensity domain. this paper, we define a novel burst detection problem to accurately of attributes. For formulate two-class classification approach isolate...