2026-10-12T23:26:41Z https://www.scidb.cn/oai

10.57760/sciencedb.j00001.01868 2026-08-03T15:52:35Z

2026-08-03 Multimodal Image and Acoustic Dataset of Paulownia Guqin Soundboards doi:10.57760/sciencedb.j00001.01868 en This dataset was collected in 2024 in the semi-anechoic chamber of the Department of Music Science and Technology, Shenyang Conservatory of Music, Shenyang, Liaoning Province, China. The dataset contains 16 kiln-dried Paulownia guqin soundboards from the same batch, indexed from Q001 to Q016. All samples were in the initial hollowed white-wood stage and had not been coated with lacquer or ash ground. A robotic arm was used to perform standardized mo-style string excitation, and a twelve-channel microphone array was used to synchronously record acoustic responses. Soundboard images, ROI images, raw acoustic engineering files and a field acquisition video were also recorded.The dataset contains 74 files with a total volume of approximately 383.66 MB. It consists of four main parts: (1) image data, including 16 raw JPG images and 16 ROI PNG images with alpha channels; (2) acoustic data, including 16 twelve-channel WAV files recorded at 48 kHz and 16-bit PCM; (3) engineering and field records, including 16 raw HEAD acoustics datafile format files and one MOV video; and (4) feature, quality-control and metadata files, including a 147-dimensional image feature matrix, a 1,074-dimensional acoustic feature matrix, a channel-level quality-control matrix, sample metadata, a data dictionary, a file manifest and a Q001 reuse workflow example.All tabular data use qin_id as the cross-modal primary key, allowing images, audio recordings, feature matrices and quality-control labels to be linked consistently. The file_manifest.csv file records file paths, file sizes, SHA-256 checksums and basic format metadata. The sample_metadata.csv file describes sample material, processing stage, acquisition environment, recommended channel and sampling rate. The channel_qc_label_matrix.csv file provides quality labels for the 12 acoustic channels, including Usable, Clipped, Low SNR, Low Signal and Auxiliary, to support channel selection. The main files for all Q001-Q016 samples are complete, with no sample-level missing data. The HDF files are HEAD acoustics datafile format files and should be read with software compatible with HEAD acoustics data files. Guqin; Paulownia soundboard; image texture; acoustic array; robotic excitation; multimodal dataset cao ye Li Yuanchen LYU Shiliang PUBLIC https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank