@inproceedings{maazsaeed2026_117980,
  title = {Comparison of Vision-based Object Detectors and an Audio Transformer for Whale Sound Event Detection},
  author = {Maaz Saeed and Bidarouni, Amir Latifi and Lukashevich, Hanna and Abeßer, Jakob},
  year = {2026},
  booktitle = {Proceedings of the 11th Workshop on Detection and Classication of Acoustic Scenes and Events (DCASE 2026)},
  pages = {176-180},
  publisher = {Zenodo},
  isbn = {979-8-234-25461-0},
  doi = {10.5281/zenodo.22954971},
  url = {https://fis.uni-bamberg.de/handle/uniba/117980},
  language = {eng}
}

@incollection{gerhardt2026_115155,
  title = {Audio-Based Tools for Decontextualisation Detection},
  author = {Gerhardt, Milica and Cuccovillo, Luca and Aichroth, Patrick and Abeßer, Jakob},
  year = {2026},
  booktitle = {Countering Disinformation in the Era of Generative AI},
  pages = {317-341},
  publisher = {Springer Nature Switzerland},
  address = {Cham},
  isbn = {9783032117816},
  doi = {10.1007/978-3-032-11782-3\_11},
  url = {https://fis.uni-bamberg.de/handle/uniba/115155},
  language = {eng}
}

@article{balke2026_112997,
  title = {JSD : A Dataset for Structure Analysis in Jazz Music},
  author = {Balke, Stefan and Reck, Julian and Weiß, Christof and Abeßer, Jakob and Müller, Meinard},
  year = {2026},
  pages = {156-172},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-112997x},
  url = {https://fis.uni-bamberg.de/handle/uniba/112997},
  language = {eng},
  abstract = {Given a music recording, music structure analysis aims at identifying important structural elements and segmenting the recording according to these elements. In jazz music, a performance is often structured by repeating harmonic schemata (known as choruses), which lay the foundation for improvisation by soloists. Within the fields of music information retrieval (MIR) and computational musicology, the Weimar Jazz Database (WJD) has turned out to be an extremely valuable resource for jazz research. Containing high-quality solo transcriptions for 456 solo sections, the dataset opened up new avenues for the understanding of creative processes in jazz improvisation using computational methods. In this paper, we complement this dataset by introducing the Jazz Structure Dataset (JSD), which provides annotations on structure and instrumentation of entire recordings. The JSD comprises 340 recordings with more than 3000 annotated segments, along with a segment-wise encoding of the solo and accompanying instruments. These annotations provide the basis for training, testing, and evaluating models for various important MIR tasks, including structure analysis, solo detection, or instrument recognition. As an example application, we consider the task of structure boundary detection. Based on a traditional novelty-based as well as a more recent data-driven approach using deep learning, we indicate the potential of the JSD while critically reflecting on some evaluation aspects of structure analysis. In this context, we also demonstrate how the JSD annotations and analysis results can be made accessible in a user-friendly way via web-based interfaces for data inspection and visualization. All annotations, experimental results, and code for reproducibility are made publicly available for research purposes.},
  keywords = {Dataset, Music Structure Analysis, Boundary Detection, Instrument Recognition, Jazz}
}

@inproceedings{nadar2026_112996,
  title = {Towards Interpreting and Improving the Latent Space for Musical Chord Recognition},
  author = {Nadar, Christon R. and Taenzer, Michael and Abeßer, Jakob},
  year = {2026},
  pages = {74-79},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-112996x},
  url = {https://fis.uni-bamberg.de/handle/uniba/112996},
  language = {eng},
  abstract = {Automatic chord recognition (ACR) naturally faces musical ambiguities between chord classes. These can be responsible for many misclassifications, especially in large chord vocabularies. In this paper, we propose a metric learning approach utilizing a triplet loss for the task of ACR in order to reduce chord ambiguities. In particular, we investigate how metric learning with different triplet sampling strategies re-aligns the distances between different chord classes in the latent space. Our main finding is that metric learning significantly improves the ACR performance for two taxonomies with five and nine chord classes.},
  keywords = {Musical Chord Recognition}
}

@inproceedings{lukashevich2026_112894,
  title = {How Reliable are Posterior Class Probabilties in Automatic Music Classification?},
  author = {Lukashevich, Hanna and Grollmisch, Sascha and Abeßer, Jakob and Stober, Sebastian and Bös, Joachim},
  year = {2026},
  pages = {45-50},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-112894x},
  url = {https://fis.uni-bamberg.de/handle/uniba/112894},
  language = {eng},
  abstract = {Music classification algorithms use signal processing and machine learning approaches to extract and enrich metadata for audio recordings in music archives. Common tasks include music genre classification, where each song is assigned a single label (such as Rock, Pop, or Jazz), and musical instrument classification. Since music metadata can be ambiguous, classification algorithms cannot always achieve fully accurate predictions. Therefore, our focus extends beyond the correctly estimated class labels to include realistic confidence values for each potential genre or instrument label. In practice, many state-of-the-art classification algorithms based on deep neural networks exhibit overconfident predictions, complicating the interpretation of the final output values. In this work, we examine whether the issue of overconfident predictions and, consequently, non-representative confidence values is also relevant to music genre classification and musical instrument classification. Moreover, we describe techniques to mitigate this behavior and assess the impact of deep ensembles and temperature scaling in generating more realistic confidence outputs, which can be directly employed in real-world music tagging applications.},
  keywords = {music information retrieval, music classifcation, uncertainty, temperature scaling, deep ensembles}
}

@article{abeer2026_112893,
  title = {How Robust are Audio Embeddings for Polyphonic Sound Event Tagging?},
  author = {Abeßer, Jakob and Grollmisch, Sascha and Müller, Meinard},
  year = {2026},
  pages = {2658-2667},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-112893x},
  url = {https://fis.uni-bamberg.de/handle/uniba/112893},
  language = {eng},
  abstract = {Sound classification algorithms are challenged by the natural variability of everyday sounds, particularly for large sound class taxonomies. In order to be applicable in real-life environments, such algorithms must also be able to handle polyphonic scenarios, where simultaneously occurring and overlapping sound events need to be classified. With the rapid progress of deep learning, several deep audio embeddings (DAEs) have been proposed as pre-trained feature representations for sound classification. In this article, we analyze the embedding spaces of two non-trainable audio representations (NTARs) and five DAEs for sound classification in polyphonic scenarios (sound event tagging) and make several contributions. First, we compare general properties like the inter-correlation between feature dimensions and the scattering of sound classes in the embedding spaces. Second, we test the robustness of the embeddings against several audio degradations and propose two sensitivity measures based on a class-agnostic and a class-centric view on the resulting drift in the embedding space. Finally, as a central contribution, we study how a blending between pairs of sounds maps to embedding space trajectories and how the path of these trajectories can cause classification errors due to their proximity to other sound classes. Throughout our analyses, the PANN embeddings have shown the best overall performance for low-polyphony sound event tagging.},
  keywords = {sound event tagging, sound polyphony, deep audio embeddings, embedding space}
}

@inproceedings{he2026_112891,
  title = {Enhancing Multiscale Features for Efficient Acoustic Scene Classification with One-Dimensional Separate CNN},
  author = {He, Yuxuan and Raake, Alexander and Abeßer, Jakob},
  year = {2026},
  pages = {125-129},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-112891x},
  url = {https://fis.uni-bamberg.de/handle/uniba/112891},
  language = {eng},
  abstract = {Acoustic Scene Classifcation (ASC) is a fundamental task in audio signal processing, aiming to classify the location of an environmental audio recording. Recent advances focus on improving ASC model effciency, particularly in resource-constrained environments. Convolutional neural networks (CNNs) remain the dominant approach due to their high performance, with recent focus on 1D kernels, such as in the Time-Frequency Separate Network (TFSepNet), for reducing model complexity and computational cost. However,TFSepNet performs feature extraction using a fixed receptive field in both time and frequency dimensions, which restrictsits ability to capture multiscale contextual patterns. In this study, we investigate the integration of multiscale feature extraction modules into TF-SepNet, with the aim of improving model effciency by balancing accuracy and complexity. We propose three architectures, TFSepDCD-Net, TFSepSPP-Net, andTFSepASPP-Net, each with two architectural variants based on TF-SepNet: one replaces its max pooling layers, and the other replaces its final convolutional layer. Each architecture has three confgurations corresponding to different model sizes - small, medium, and large - to explore the tradeoff between accuracy and model complexity. Our experiments show that in corporating multiscale modules allows smaller models to achieve comparable or even superior accuracy to larger baselines. These fndings highlight the potential of multiscale representations for improving the effciency of CNN-based ASC systems, especially in 1D separate architectures like TF-SepNet.},
  keywords = {acoustic scene classifcation, time-frequency separate network, 1D kernels, multiscale feature extraction}
}

@inproceedings{dilip2026_112887,
  title = {A Three-Level Evaluation Protocoll for Acoustic Scene Understanding of Large Language Audio Models},
  author = {Dilip, Harish and Abeßer, Jakob},
  year = {2026},
  pages = {185-189},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-112887x},
  url = {https://fis.uni-bamberg.de/handle/uniba/112887},
  language = {eng},
  abstract = {Reaching a semantic understanding of complex acoustic scenes requires computational models to capture the temporal-spatial sound source composition as well as individual sound events. This is a great challenge for computational models due to the large variety of everyday sound events and the extensive temporal-spectral overlap in real-life acoustic scenes. In this work, we aim to evaluate the acoustic scene understanding capabilities of two large audio-language models (LALMs). As a challenging scenario, we use the USM dataset, which features synthetic urban soundscapes with 2-6 overlapping sound sources per mixture. Our main contribution is a novel three-layer evaluation protocol, which includes four analysis tasks for low-level sound event understanding (sound event tagging), mid-level understanding and reasoning (sound polyphony estimation, sound source loudness ranking), as well as high-level scene understanding (audio captioning). We apply standardized metrics to assess the model's performances for each task. The proposed mulit-layer protocol allows for a fine-grained analysis of model behavior across soundscapes of various complexity levels. Our results indicate that despite their rmarkable controllability using textual instructions, the ability of state-of-the-art LALMs to understand acoustic scenes is still limited as the performance on individual analysis tasks desgrades with increasing sound polyphony.},
  keywords = {audio captioning, large audio-language models, sound event tagging, sound polyphony estimation, sound source loudness ranking}
}

@article{abeer2026_112771,
  title = {Sound Recurrence Analysis for Acoustic Scene Classification},
  author = {Abeßer, Jakob and Liang, Zhiwei and Seeber, Bernhard},
  year = {2026},
  pages = {1-15},
  publisher = {Otto-Friedrich-Universität},
  address = {Bamberg},
  urn = {urn:nbn:de:bvb:473-irb-112771x},
  url = {https://fis.uni-bamberg.de/handle/uniba/112771},
  language = {eng},
  abstract = {In everyday life, people experience different soundscapes in which natural sounds, animal noises, and man-made sounds blend together. Although there have been several studies on the importance of recurring sound patterns in music and language, the relevance of this phenomenon in natural soundscapes is still largely unexplored. In this article, we study the repetition patterns of harmonic and transient sound events as potential cues for acoustic scene classification (ASC). In the first part of our study, our aim is to identify acoustic scene classes that exhibit characteristic sound repetition patterns concerning harmonic and transient sounds. We propose three metrics to measure the overall prevalence of sound repetitions as well as their repetition periods and temporal stability. In the second part, we evaluate three strategies to incorporate self-similarity matrices as an additional input feature to a convolutional neural network architecture for ASC. We observe the characteristic repetition of transient sounds in recordings of “park” and “street traffic” as well as harmonic sound repetitions in acoustic scene classes related to public transportation. In the ASC experiments, hybrid network architectures, which combine spectrogram features and features from sound recurrence analysis, show increased accuracy for those classes with prominent sound repetition patterns. Our findings provide additional perspective on the distinctions among acoustic scenes previously primarily ascribed in the literature to their spectral features.},
  keywords = {Acoustic scene classification, Sound recurrence analysis, Sound repetition patterns, Self-similarity matrix, Harmonic-percussive source separation, Result fusion, Ensemble models}
}

@inproceedings{kumar2025_113068,
  title = {Improvement and Cross-Domain Evaluation of SlowFast Networks},
  author = {Kumar, Ravi and Grollmisch, Sascha and Abeßer, Jakob},
  year = {2025},
  booktitle = {2025 AES International Conference on Artificial Intelligence and Machine Learning for Audio: Queen Mary University of London 8th to 10th, September 2025 : Proceedings},
  pages = {211-219},
  url = {https://fis.uni-bamberg.de/handle/uniba/113068},
  language = {eng}
}

@inproceedings{he2025_113069,
  title = {From CNN to Reservoir Computing : A New Perspective on Acoustic Scene Classification},
  author = {He, Yuxuan and Hosseini, Alireza Molla Ali and Abeßer, Jakob and Jaurigue, Lina and Raake, Alexander and Lüdge, Kathy},
  year = {2025},
  booktitle = {2025 AES International Conference on Artificial Intelligence and Machine Learning for Audio: Queen Mary University of London 8th to 10th, September 2025 : Proceedings},
  pages = {201-210},
  url = {https://fis.uni-bamberg.de/handle/uniba/113069},
  language = {eng}
}

@inproceedings{lenk2025_111571,
  title = {NeuroSensEar : Neuromorphic Acoustic Sensing for the High-Performance Hearing Aids of Tomorrow},
  author = {Lenk, Claudia and He, Yuxuan and Hosseini, Alireza and Zheng, Fang and Mukherjee, Aayushmi and Abeßer, Jakob and Jaurigue, Lina and Lüdge, Kathy and Raake, Alexander and Ivanov, Tzvetan and Pieper, Iko and Fritzsche, Paul and Küller, Jan and Beer, Daniel and Schöneich, Stefan and Nowotny, Manuela and Uziel, Sebastian and Hutschenreuther, Tino and Ziegler, Martin},
  year = {2025},
  booktitle = {Proceedings of DAS/DAGA 2025 : 51st Annual Meeting on Acoustics : March 17-20, 2025, Copenhagen},
  pages = {509-512},
  address = {Berlin},
  isbn = {978-3-939296-23-2},
  doi = {10.71568/dasdaga2025.523},
  url = {https://fis.uni-bamberg.de/handle/uniba/111571},
  language = {eng}
}

@inproceedings{he2025_111568,
  title = {Enhancing Multiscale Features for Efficient Acoustic Scene Classification with One-Dimensional Separate CNN},
  author = {He, Yuxuan and Raake, Alexander and Abeßer, Jakob},
  year = {2025},
  booktitle = {Proceedings of the 10th Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE 2025), October 2025},
  pages = {125-129},
  publisher = {Zenodo},
  isbn = {978-84-09-77652-8},
  doi = {10.5281/zenodo.17251589},
  url = {https://fis.uni-bamberg.de/handle/uniba/111568},
  language = {eng},
  abstract = {Acoustic Scene Classifcation (ASC) is a fundamental task in audio signal processing, aiming to classify the location of an environmental audio recording. Recent advances focus on improving ASC model effciency, particularly in resource-constrained environments. Convolutional neural networks (CNNs) remain the dominant approach due to their high performance, with recent focus on 1D kernels, such as in the Time-Frequency Separate Network (TFSepNet), for reducing model complexity and computational cost. However,TFSepNet performs feature extraction using a fixed receptive field in both time and frequency dimensions, which restrictsits ability to capture multiscale contextual patterns. In this study, we investigate the integration of multiscale feature extraction modules into TF-SepNet, with the aim of improving model effciency by balancing accuracy and complexity. We propose three architectures, TFSepDCD-Net, TFSepSPP-Net, andTFSepASPP-Net, each with two architectural variants based on TF-SepNet: one replaces its max pooling layers, and the other replaces its final convolutional layer. Each architecture has three confgurations corresponding to different model sizes - small, medium, and large - to explore the tradeoff between accuracy and model complexity. Our experiments show that in corporating multiscale modules allows smaller models to achieve comparable or even superior accuracy to larger baselines. These fndings highlight the potential of multiscale representations for improving the effciency of CNN-based ASC systems, especially in 1D separate architectures like TF-SepNet.},
  keywords = {acoustic scene classifcation, time-frequency separate network, 1D kernels, multiscale feature extraction}
}

@inproceedings{dilip2025_111565,
  title = {A Three-Level Evaluation Protocoll for Acoustic Scene Understanding of Large Language Audio Models},
  author = {Dilip, Harish and Abeßer, Jakob},
  year = {2025},
  booktitle = {Proceedings of the 10th Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE 2025), October 2025},
  pages = {185-189},
  publisher = {zenodo},
  isbn = {978-84-09-77652-8},
  doi = {10.5281/zenodo.17251589},
  url = {https://fis.uni-bamberg.de/handle/uniba/111565},
  language = {eng},
  abstract = {Reaching a semantic understanding of complex acoustic scenes requires computational models to capture the temporal-spatial sound source composition as well as individual sound events. This is a great challenge for computational models due to the large variety of everyday sound events and the extensive temporal-spectral overlap in real-life acoustic scenes. In this work, we aim to evaluate the acoustic scene understanding capabilities of two large audio-language models (LALMs). As a challenging scenario, we use the USM dataset, which features synthetic urban soundscapes with 2-6 overlapping sound sources per mixture. Our main contribution is a novel three-layer evaluation protocol, which includes four analysis tasks for low-level sound event understanding (sound event tagging), mid-level understanding and reasoning (sound polyphony estimation, sound source loudness ranking), as well as high-level scene understanding (audio captioning). We apply standardized metrics to assess the model's performances for each task. The proposed mulit-layer protocol allows for a fine-grained analysis of model behavior across soundscapes of various complexity levels. Our results indicate that despite their rmarkable controllability using textual instructions, the ability of state-of-the-art LALMs to understand acoustic scenes is still limited as the performance on individual analysis tasks desgrades with increasing sound polyphony.},
  keywords = {audio captioning, large audio-language models, sound event tagging, sound polyphony estimation, sound source loudness ranking}
}

@inproceedings{grollmisch2025_111564,
  title = {Semi-Supervised Learning for Acoustic Scene Classification using FixMatch},
  author = {Grollmisch, Sascha and Kumar, Ravi and Abeßer, Jakob},
  year = {2025},
  booktitle = {Proceedings of DAS/DAGA 2025 : 51st Annual Meeting on Acoustics : March 17-20, 2025, Copenhagen},
  pages = {109-111},
  address = {Berlin},
  isbn = {978-3-939296-23-2},
  url = {https://fis.uni-bamberg.de/handle/uniba/111564},
  language = {eng}
}

@article{abeer2025_111563,
  title = {Fortschritte in der automatischen Erkennung von Vogelstimmen},
  author = {Abeßer, Jakob and Lukashevich, Hanna and Ziegler, Sebastian and Bös, Joachim},
  year = {2025},
  journal = {Akustik Journal},
  number = {3},
  pages = {7-16},
  address = {Berlin},
  issn = {2569–1597},
  url = {https://fis.uni-bamberg.de/handle/uniba/111563},
  language = {deu}
}

@article{abeer2025_111561,
  title = {Sound Recurrence Analysis for Acoustic Scene Classification},
  author = {Abeßer, Jakob and Liang, Zhiwei and Seeber, Bernhard},
  year = {2025},
  journal = {EURASIP Journal on audio, speech, and music processing : EURASIP JASMP},
  volume = {1},
  number = {1},
  pages = {1-15},
  publisher = {Springer},
  address = {Heidelberg},
  issn = {1687-4722},
  doi = {10.1186/s13636-024-00390-2},
  url = {https://fis.uni-bamberg.de/handle/uniba/111561},
  language = {eng},
  abstract = {In everyday life, people experience different soundscapes in which natural sounds, animal noises, and man-made sounds blend together. Although there have been several studies on the importance of recurring sound patterns in music and language, the relevance of this phenomenon in natural soundscapes is still largely unexplored. In this article, we study the repetition patterns of harmonic and transient sound events as potential cues for acoustic scene classification (ASC). In the first part of our study, our aim is to identify acoustic scene classes that exhibit characteristic sound repetition patterns concerning harmonic and transient sounds. We propose three metrics to measure the overall prevalence of sound repetitions as well as their repetition periods and temporal stability. In the second part, we evaluate three strategies to incorporate self-similarity matrices as an additional input feature to a convolutional neural network architecture for ASC. We observe the characteristic repetition of transient sounds in recordings of “park” and “street traffic” as well as harmonic sound repetitions in acoustic scene classes related to public transportation. In the ASC experiments, hybrid network architectures, which combine spectrogram features and features from sound recurrence analysis, show increased accuracy for those classes with prominent sound repetition patterns. Our findings provide additional perspective on the distinctions among acoustic scenes previously primarily ascribed in the literature to their spectral features.},
  keywords = {Acoustic scene classification, Sound recurrence analysis, Sound repetition patterns, Self-similarity matrix, Harmonic-percussive source separation, Result fusion, Ensemble models}
}

@inproceedings{ngamthipwatthana2024_111590,
  title = {Towards Measuring and Forecasting Noise Exposure at the VELTINS-Arena in Gelsenkirchen, Germany},
  author = {Ngamthipwatthana, Pitchapa and Götze, Marco and Kátai, András and Abeßer, Jakob},
  year = {2024},
  booktitle = {2024 IEEE 5th International Symposium on the Internet of Sounds (IS2) : proceedings},
  pages = {137-144},
  publisher = {IEEE},
  doi = {10.1109/IS262782.2024.10704088},
  url = {https://fis.uni-bamberg.de/handle/uniba/111590},
  language = {eng}
}

@inproceedings{grollmisch2024_111589,
  title = {Selbstüberwachtes Vortraining zur Verbesserung automatischer Audioklassifikationsalgorithmen},
  author = {Grollmisch, Sascha and Abeßer, Jakob and Bös, Joachim},
  year = {2024},
  booktitle = {Fortschritte der Akustik - DAGA 2024". Tagungsband - Proceedings: DAGA 2024 - 50. Jahrestagung für Akustik, 18.-21. März 2024, Hannover},
  pages = {1406-1409},
  address = {Berlin},
  isbn = {978-3-939296-22-5},
  url = {https://fis.uni-bamberg.de/handle/uniba/111589},
  language = {deu}
}

@inproceedings{latifibidarouni2024_111587,
  title = {Towards Domain Shift in Location-Mismatch Scenarios for Bird Activity Detection},
  author = {Latifi Bidarouni, Amir and Abeßer, Jakob},
  year = {2024},
  booktitle = {32nd European Signal Processing Conference (EUSIPCO 2024) : proceedings},
  pages = {1267-1271},
  publisher = {IEEE},
  address = {Piscataway, NJ},
  isbn = {978-94-645936-1-7},
  doi = {10.23919/EUSIPCO63174.2024.10715313},
  url = {https://fis.uni-bamberg.de/handle/uniba/111587},
  language = {eng}
}

@inproceedings{apostolidis2024_111586,
  title = {Visual and Audio Scene Classification for Detecting Discrepancies in Video : A Baseline Method and Experimental Protocol},
  author = {Apostolidis, Konstantinos and Abeßer, Jakob and Cuccovillo, Luca and Mezaris, Vasileios},
  year = {2024},
  booktitle = {MAD '24: Proceedings of the 3rd ACM International Workshop on Multimedia AI against Disinformation},
  pages = {30-36},
  publisher = {New York, NYUnited States},
  address = {Association for Computing Machinery},
  isbn = {979-8-4007-0552-6},
  doi = {10.1145/3643491.3660287},
  url = {https://fis.uni-bamberg.de/handle/uniba/111586},
  language = {eng}
}

@inproceedings{lukashevich2023_111597,
  title = {How Reliable are Posterior Class Probabilties in Automatic Music Classification?},
  author = {Lukashevich, Hanna and Grollmisch, Sascha and Abeßer, Jakob and Stober, Sebastian and Bös, Joachim},
  year = {2023},
  booktitle = {AM '23: Proceedings of the 18th International Audio Mostly Conference},
  pages = {45-50},
  publisher = {ACM},
  isbn = {979-8-4007-0818-3},
  doi = {10.1145/3616195.3616228},
  url = {https://fis.uni-bamberg.de/handle/uniba/111597},
  language = {eng},
  abstract = {Music classification algorithms use signal processing and machine learning approaches to extract and enrich metadata for audio recordings in music archives. Common tasks include music genre classification, where each song is assigned a single label (such as Rock, Pop, or Jazz), and musical instrument classification. Since music metadata can be ambiguous, classification algorithms cannot always achieve fully accurate predictions. Therefore, our focus extends beyond the correctly estimated class labels to include realistic confidence values for each potential genre or instrument label. In practice, many state-of-the-art classification algorithms based on deep neural networks exhibit overconfident predictions, complicating the interpretation of the final output values. In this work, we examine whether the issue of overconfident predictions and, consequently, non-representative confidence values is also relevant to music genre classification and musical instrument classification. Moreover, we describe techniques to mitigate this behavior and assess the impact of deep ensembles and temperature scaling in generating more realistic confidence outputs, which can be directly employed in real-world music tagging applications.},
  keywords = {music information retrieval, music classifcation, uncertainty, temperature scaling, deep ensembles}
}

@inproceedings{lukashevich2023_111596,
  title = {Quantifying Uncertainty in Music Genre Classification},
  author = {Lukashevich, Hanna and Grollmisch, Sascha and Abeßer, Jakob},
  year = {2023},
  booktitle = {Tagungsband, DAGA 2023 - 49. Jahrestagung für Akustik : 06.-09. März 2023, Hamburg},
  pages = {1378-1381},
  address = {Berlin},
  isbn = {978-3-939296-21-8},
  url = {https://fis.uni-bamberg.de/handle/uniba/111596},
  language = {eng}
}

@inproceedings{grollmisch2023_111594,
  title = {Uncertainty in Semi-Supervised Audio Classification : A Novel Extension for FixMatch},
  author = {Grollmisch, Sascha and Cano, Estefanía and Lukashevich, Hanna and Abeßer, Jakob},
  year = {2023},
  booktitle = {31st European Signal Processing Conference (EUSIPCO 2024) : proceedings},
  pages = {161-165},
  publisher = {IEEE},
  address = {Piscataway, NJ},
  doi = {10.23919/eusipco58844.2023.10289789},
  url = {https://fis.uni-bamberg.de/handle/uniba/111594},
  language = {eng}
}

@inproceedings{bidarouni2023_111593,
  title = {Unsupervised Feature-Space Domain Adaptation applied for Audio Classification},
  author = {Bidarouni, Amir Latifi and Abeßer, Jakob},
  year = {2023},
  booktitle = {4th International Symposium on the Internet of Sounds : Congress Center Le Benedettine, Pisa, 26-27 October 2023},
  pages = {1-7},
  publisher = {IEEE},
  isbn = {979-8-3503-8254-9},
  doi = {10.1109/ieeeconf59510.2023.10335455},
  url = {https://fis.uni-bamberg.de/handle/uniba/111593},
  language = {eng}
}

@article{abeer2023_111592,
  title = {Human and Machine Performance in Counting Sound Classes in Single-Channel Soundscapes},
  author = {Abeßer, Jakob and Ullah, Asad and Ziegler, Sebastian and Grollmisch, Sascha},
  year = {2023},
  journal = {Journal of the Audio Engineering Society},
  volume = {71},
  number = {12},
  pages = {860-872},
  publisher = {AES},
  address = {New York, NY},
  issn = {1549-4950},
  url = {https://fis.uni-bamberg.de/handle/uniba/111592},
  language = {eng}
}

@article{abeer2023_111591,
  title = {How Robust are Audio Embeddings for Polyphonic Sound Event Tagging?},
  author = {Abeßer, Jakob and Grollmisch, Sascha and Müller, Meinard},
  year = {2023},
  journal = {IEEE ACM transactions on audio, speech, and language processing : TASLP},
  volume = {31},
  pages = {2658-2667},
  publisher = {IEEE},
  address = {New York, NY},
  issn = {2329-9290},
  doi = {10.1109/taslp.2023.3293032},
  url = {https://fis.uni-bamberg.de/handle/uniba/111591},
  language = {eng},
  abstract = {Sound classification algorithms are challenged by the natural variability of everyday sounds, particularly for large sound class taxonomies. In order to be applicable in real-life environments, such algorithms must also be able to handle polyphonic scenarios, where simultaneously occurring and overlapping sound events need to be classified. With the rapid progress of deep learning, several deep audio embeddings (DAEs) have been proposed as pre-trained feature representations for sound classification. In this article, we analyze the embedding spaces of two non-trainable audio representations (NTARs) and five DAEs for sound classification in polyphonic scenarios (sound event tagging) and make several contributions. First, we compare general properties like the inter-correlation between feature dimensions and the scattering of sound classes in the embedding spaces. Second, we test the robustness of the embeddings against several audio degradations and propose two sensitivity measures based on a class-agnostic and a class-centric view on the resulting drift in the embedding space. Finally, as a central contribution, we study how a blending between pairs of sounds maps to embedding space trajectories and how the path of these trajectories can cause classification errors due to their proximity to other sound classes. Throughout our analyses, the PANN embeddings have shown the best overall performance for low-polyphony sound event tagging.},
  keywords = {sound event tagging, sound polyphony, deep audio embeddings, embedding space}
}

@inproceedings{abeer2022_111761,
  title = {Construction-sAIt : Multi-modal AI-driven technologies for construction site monitoring},
  author = {Abeßer, Jakob and Loos, Alexander and Sharma, Prachi},
  year = {2022},
  booktitle = {Tagungsband, DAGA 2022 - 48. Jahrestagung für Akustik : 21.-24. März 2022, Stuttgart und Online},
  pages = {90-93},
  address = {Berlin},
  isbn = {978-3-939296-20-1},
  url = {https://fis.uni-bamberg.de/handle/uniba/111761},
  language = {eng}
}

@inproceedings{abeer2022_111762,
  title = {Analyzing Bird and Bat Activity in Agricultural Environments using AI-driven Audio Monitoring},
  author = {Abeßer, Jakob and Wang, Xiaoyi and Bänsch, Svenja and Scherber, Christoph and Lukashevich, Hanna},
  year = {2022},
  booktitle = {Fortschritte der Akustik - DAGA 2022 : 48. Jahrestagung für Akustik, 21. - 24. März 2022, Stuttgart und Online},
  pages = {123-126},
  address = {Berlin},
  isbn = {978-3-939296-20-1},
  url = {https://fis.uni-bamberg.de/handle/uniba/111762},
  language = {eng}
}

@article{balke2022_111763,
  title = {JSD : A Dataset for Structure Analysis in Jazz Music},
  author = {Balke, Stefan and Reck, Julian and Weiß, Christof and Abeßer, Jakob and Müller, Meinard},
  year = {2022},
  journal = {Transactions of the International Society for Music Information Retrieval},
  volume = {5},
  number = {1},
  pages = {156-172},
  publisher = {Ubiquity Press, Ltd.},
  address = {London},
  issn = {2514-3298},
  doi = {10.5334/tismir.131},
  url = {https://fis.uni-bamberg.de/handle/uniba/111763},
  language = {eng},
  abstract = {Given a music recording, music structure analysis aims at identifying important structural elements and segmenting the recording according to these elements. In jazz music, a performance is often structured by repeating harmonic schemata (known as choruses), which lay the foundation for improvisation by soloists. Within the fields of music information retrieval (MIR) and computational musicology, the Weimar Jazz Database (WJD) has turned out to be an extremely valuable resource for jazz research. Containing high-quality solo transcriptions for 456 solo sections, the dataset opened up new avenues for the understanding of creative processes in jazz improvisation using computational methods. In this paper, we complement this dataset by introducing the Jazz Structure Dataset (JSD), which provides annotations on structure and instrumentation of entire recordings. The JSD comprises 340 recordings with more than 3000 annotated segments, along with a segment-wise encoding of the solo and accompanying instruments. These annotations provide the basis for training, testing, and evaluating models for various important MIR tasks, including structure analysis, solo detection, or instrument recognition. As an example application, we consider the task of structure boundary detection. Based on a traditional novelty-based as well as a more recent data-driven approach using deep learning, we indicate the potential of the JSD while critically reflecting on some evaluation aspects of structure analysis. In this context, we also demonstrate how the JSD annotations and analysis results can be made accessible in a user-friendly way via web-based interfaces for data inspection and visualization. All annotations, experimental results, and code for reproducibility are made publicly available for research purposes.},
  keywords = {Dataset, Music Structure Analysis, Boundary Detection, Instrument Recognition, Jazz}
}

@inproceedings{nadar2022_111765,
  title = {Towards Interpreting and Improving the Latent Space for Musical Chord Recognition},
  author = {Nadar, Christon R. and Taenzer, Michael and Abeßer, Jakob},
  year = {2022},
  booktitle = {Standing wave : ICMC 2022 : International Computer Music Conference, University of Limerick, Ireland, 2022},
  pages = {74-79},
  publisher = {International Computer Music Association, Inc.},
  address = {San Francisco, California, USA},
  isbn = {978-0-9713192-6-4},
  url = {https://fis.uni-bamberg.de/handle/uniba/111765},
  language = {eng},
  abstract = {Automatic chord recognition (ACR) naturally faces musical ambiguities between chord classes. These can be responsible for many misclassifications, especially in large chord vocabularies. In this paper, we propose a metric learning approach utilizing a triplet loss for the task of ACR in order to reduce chord ambiguities. In particular, we investigate how metric learning with different triplet sampling strategies re-aligns the distances between different chord classes in the latent space. Our main finding is that metric learning significantly improves the ACR performance for two taxonomies with five and nine chord classes.},
  keywords = {Musical Chord Recognition}
}

@inproceedings{ribecky2022_111766,
  title = {Multi-Input Architecture and Disentangled Representation Learning for Multi-Dimensional Modeling of Music Similarity},
  author = {Ribecky, Sebastian and Lukashevich, Hanna and Abeßer, Jakob},
  year = {2022},
  publisher = {AES},
  address = {New York, NY},
  url = {https://fis.uni-bamberg.de/handle/uniba/111766},
  language = {eng}
}

@inproceedings{abeer2022_111760,
  title = {Classifying Sounds in Polyphonic Urban Sound Scenes},
  author = {Abeßer, Jakob},
  year = {2022},
  publisher = {AES},
  address = {New York, NY},
  url = {https://fis.uni-bamberg.de/handle/uniba/111760},
  language = {eng}
}

@inproceedings{gourishetti2022_111764,
  title = {Potentials and Challenges of AI-based Audio Analysis in Industrial Sound Analysis},
  author = {Gourishetti, Saichand and Grollmisch, Sascha and Abeßer, Jakob and Liebetrau, Judith},
  year = {2022},
  booktitle = {Tagungsband, DAGA 2022 - 48. Jahrestagung für Akustik : 21.-24. März 2022, Stuttgart und Online},
  pages = {94-97},
  address = {Berlin},
  isbn = {978-3-939296-20-1},
  url = {https://fis.uni-bamberg.de/handle/uniba/111764},
  language = {eng}
}

