Tonal Features of DataSEC Sounds: Single, Double, and Triple-Band Tonality Distributions

This dataset supports the tonal analysis presented in the associated study and is based on the DataSEC dataset (DOI: 10.5281/zenodo.15393250), a previously developed and validated collection by the authors. DataSEC includes 5,048 .wav audio recordings covering 22 environmental sound classes, further subdivided into a total of 40 sub-classes. The present dataset was created using a rule-based method designed by the authors to detect tonal structures in third-octave band spectra. The aim is to support the development of spectral smoothing strategies in environmental sound analysis. The detection approach identifies tonal components based on the prominence of spectral peaks, without relying on psychoacoustic models, making it suitable for scalable integration into machine learning pipelines. Tonal analysis was performed on all 5,048 audio files of the DataSEC dataset. A third-octave band was labeled as tonal if its sound pressure level (SPL) exceeded the SPLs of its adjacent bands by at least 5 dB. If no single-band tonal peaks were found, the algorithm proceeded to detect dual-band and triple-band tonalities, defined as follows: Dual Tonality: Two adjacent bands, where the first exceeds the band to its left and the second exceeds the band to its right, each by at least 5 dB. Triple Tonality: Three consecutive bands, where the first and third exceed their respective outer neighbors by at least 5 dB. Each audio file was assigned only one tonality type—single, dual, triple, or non-tonal—based on a hierarchical evaluation. Files classified as single-tonal were not further evaluated for dual or triple tonalities, ensuring mutually exclusive labeling. The Dataset is organized into IP archives: "Tonality_Analysis.zip": This folder contains figure files for each audio recording, indicating whether the file is single-tonal, dual-tonal, triple-tonal, or non-tonal. These visualizations support verification and qualitative inspection of the detection process. "Tonality_Probabilities.zip" :This folder includes probability plots for each of the 40 sound classes. Each plot shows the likelihood (as a proportion of recordings) of observing single-, dual-, or triple-band tonal components at each third-octave band center frequency. This probabilistic view supports the identification of class-specific tonal patterns and informs tailored post-processing strategies. All files are provided in .svg vector format to ensure maximum clarity and scalability. File names directly reference the corresponding .wav recordings in the original DataSEC dataset, allowing seamless cross-referencing and traceability. This dataset enables a structured and reproducible approach to analyzing tonal content in environmental sound recordings and supports the development of automated spectral smoothing and classification tools for noise monitoring and soundscape assessment. The newer version contains updated figures in terms of visibility and readability.