Overview

Audio classification studies how to assign sounds to labels, from spoken words and music genres to bird calls, machine faults, and respiratory conditions. Systems commonly use spectrograms, mel-frequency cepstral coefficients, or raw waveforms with convolutional networks, transformers, and pretrained or self-supervised audio encoders; limited-label settings also use transfer learning, few-shot methods, and audio-language models. Benchmarks such as AudioSet, ESC-50, UrbanSound8K, and ICBHI test different tasks, but strong benchmark scores may not carry over across recording devices, locations, speakers, or species. A central challenge is building reliable models from scarce or weak labels while avoiding data leakage and shortcuts, and keeping models efficient enough for clinical and on-device use.

Latest papers

CardsList

154 papers12 papers in the last four weeks, down 29% on the four weeks before. 0.1% of all new papers.

  1. Localized time-frequency representation learning for bioacoustic classification in complex soundscapes

    Feb 19, 2025Simen Hexeberg, Mandar Chitre, Matthias Hoffmann-Kuhnt +1BioacousticsAudio Classification

  2. Classifying bioacoustic data without individual call annotations using temporal convolutional networks and feature extractors

    Oct 22, 2024Laia Garrobé Fonollosa, Douglas Gillespie, Lina Stankovic +2BioacousticsPassive Acoustic Monitoring

  3. Topology-enhanced machine learning for speech signal processing

    Nov 26, 2023Pingyao Feng, Qingrui Qu, Haiyu Zhang +4Interpretable MLSpeech Processing

  4. Robust Wake-Up Word Detection by Two-stage Multi-resolution Ensembles

    Oct 17, 2023Fernando López, Jordi Luque, Carlos Segura +1Keyword SpottingEnsemble Learning