Lion City Soundscapes (LCS)

This repository contains audio-visual recordings made at the 62 locations identified by the Singapore Soundscape Site Selection Survey, as 1-minute long audio and video files. The 1-minute long files in this repository have been cropped from much longer full-length files that were initially recorded from the respective locations as "representative" sections of recordings of those locations, and are more suitable for short-term analysis of those soundscapes. A model pre-trained on the ARAUS dataset was used to crop the 1-minute sections by identifying the sections best corresponding to its ISO12913-3:2019 perceptual quadrant in the Singapore Soundscape Site Selection Survey. For further details of the exact method, please refer to our publication: Kenneth Ooi, Jessie Goh, Hao-Weng Lin, Zhen-Ting Ong, Trevor Wong, Karn Watcharasupat, Bhan Lam, Woon-Seng Gan, Lion City Soundscapes: Modified Partitioning Around Medoids for a Perceptually Diverse Dataset of Singaporean Soundscapes, JASA Express Letters, 2024 [under review]. The original full-length files have different lengths depending on the specific recording location and range from 9-38 minutes in length, which might be more suitable for medium-term analysis of the same soundscapes. Due to memory and bandwidth limitations, only a selection of full-length files are currently accessible on this repository (the full-length binaural microphone recordings for all locations, as well as all full-length files for locations S0001, S0002, and S0041). However, preview videos of the files have been made available on YouTube (respective URLs under the youtube_link column of metadata.csv, available in this repository and also the project GitHub repository https://github.com/ntudsp/lion-city-soundscapes) and all raw files are available upon request. If you would like to download the full-length files, please ensure that your system is able to handle a download of about 1.5TB and contact us regarding the request with the email header "[LCS] Download of full-length files" for the steps to download. The following types of files are available for each individual location (all are synchronised with each other): 19-channel wave file recorded using a ZYLIA ZM-1 3rd-order Ambisonic microphone 2-channel stereo wave file generated from the 19-channel wave file upon recording using a ZYLIA ZR-1 Portable Recorder 3-channel HDF file recorded using a B&K Type 4101-B binaural microphone (first two channels) and GRAS 146AE Free-field Microphone (last channel) connected to a HEAD Acoustics SQobold Data Acquisition System. Note that "HDF" here stands for "HEAD Data Format", and not the more standard "Hierarchical Data Format". This file can be opened with the use of software such as HEAD Companion and ArtemiS 2-channel wave file recorded using the binaural microphone, as generated from the first two channels of the HDF file 1-channel wave file recorded using the GRAS 146AE Free-field Microphone, as generated from the last channel of the HDF file 360-degree video file (in spherical, MP4 format) recorded using an Insta360 One R camera and onboard microphone. Note that no windscreen was used for the onboard microphone and the audio quality is comparatively low, so we advise coupling the audio from the binaural microphone or 3rd-order Ambisonic microphone to the video if audio-visual reproduction is desired.