BioSEN: A Bio-acoustic Signal Enhancement Network for Animal Vocalizations
Most work in audio enhancement targets human speech, while bioacoustics is less studied due to noisy recordings and the distinct traits of animal sounds. To fill this gap, we adapt speech enhancement methods and build BioSEN, a model made for bioacoustic signals. BioSEN has three modules: a multi-scale dual-axis attention unit for time-frequency feature extraction, a bio-harmonic multi-scale enhancement unit for capturing harmonic structures, and an energy-adaptive gating connection unit that uses frequency weights to keep vocalizations from being removed as noise. Tests on three bioacoustic datasets show that BioSEN matches or exceeds state-of-the-art speech enhancement models while using far less computation. These results show BioSEN's strength for bioacoustic audio enhancement and its promise for biodiversity monitoring and conservation.
Code (0)
등록된 구현이 없습니다.
Tasks
Speech EnhancementSimilar Papers 제목 키워드 기반
Multi Modal Information Fusion of Acoustic and Linguistic Data for Decoding Dairy Cow Vocalizations in Animal Welfare Assessment
Understanding animal vocalizations through multi-source data fusion is crucial for assessing emotional states and enhancing animal welfare in precision livestock farming. This study aims to decode dairy cow contact calls…
Towards Leveraging Sequential Structure in Animal Vocalizations
Animal vocalizations contain sequential structures that carry important communicative information, yet most computational bioacoustics studies average the extracted frame-level features across the temporal axis, discardi…
animal2vec and MeerKAT: A self-supervised transformer for rare-event raw audio input and a large-scale reference dataset for bioacoustics
Bioacoustic research, vital for understanding animal behavior, conservation, and ecology, faces a monumental challenge: analyzing vast datasets where animal vocalizations are rare. While deep learning techniques are beco…
Audio ClassificationBenchmarkingFew-Shot LearningDecoding Poultry Vocalizations -- Natural Language Processing and Transformer Models for Semantic and Emotional Analysis
Deciphering the acoustic language of chickens offers new opportunities in animal welfare and ecological informatics. Their subtle vocal signals encode health conditions, emotional states, and dynamic interactions within …
Comparing Self-Supervised Learning Models Pre-Trained on Human Speech and Animal Vocalizations for Bioacoustics Processing
Self-supervised learning (SSL) foundation models have emerged as powerful, domain-agnostic, general-purpose feature extractors applicable to a wide range of tasks. Such models pre-trained on human speech have demonstrate…
Automatic Speech RecognitionAutomatic Speech Recognition (ASR)Self-Supervised Learningspeech-recognition+1