Sonic Inspection of Spectra for Uncovering Hidden and Historic Quasars

目前和未来的光谱测量，包括DESI、WEAVE、4MOST、MONS和PFS，将在未来十年产生数以百万计的星系光谱。此外，由综合野外单位(IFU)产生的大量大型数据立方体将分别包含数千个单独的光谱。对于这些大数据量，传统的目视检查方法正变得越来越不可行。虽然开发了机器学习(ML)工具，但仍然需要对数据进行人工数据检查，包括评估数据质量、验证拟合结果以及生成ML方法所需的训练数据集。仅靠自动化方法也限制了发现。它们被优化以识别感兴趣的已知特征，并且存在对稀有或以前未知的对象进行错误分类的风险。光谱数据的音频检查是视觉检查的一种替代方法，具有更舒适和更高效的潜力。例如，可以在外围监控发声数据，从而使其他任务能够同时执行，从而提高效率。该项目将开发发声(用声音表示数据)星系光谱的方法。最初的重点将是通过检查一维光谱来搜索红色类星体(伪装成星系)，并识别和表征光谱数据立方体中的扩展发射线区域(EELR)。后者可以帮助识别不再活跃、因此不再可见的“历史”类星体。所开发的方法可适用于从即将进行的光谱调查中提取的各种科学案例并加以调整。

Current and future spectroscopic surveys, including DESI, WEAVE, 4MOST, MOONS and PFS, will produce millions of galaxy spectra over the coming decade. Additionally, an influx of large datacubes, produced by Integral Field Units (IFUs), will each contain thousands of individual spectra. Traditional methods of visual inspection are becoming increasingly unfeasible for these large data volumes. Whilst Machine Learning (ML) tools are developed, manual data inspection is still required for tasks including, assessment of data quality, verification of fitting results, and producing the training datasets needed for ML approaches. Automated methods alone also limit discovery. They are optimised to identify known features of interest, and there is the risk of misclassification of rare or previously unknown objects. Audio inspection of spectroscopic data is an alternative method to visual inspection, with the potential to be more comfortable and efficient. For example, it may be possible to peripherally monitor sonified data, enabling other tasks to be performed simultaneously, increasing efficiency. This project will develop approaches for sonification (representing data with sound) of galaxy spectra. The initial focus will be searching for red quasars (which masquerade as galaxies) through the inspection of 1D spectra, and identifying and characterising Extended Emission Line Regions (EELRs) in spectral datacubes. The latter could help to identify 'historic' quasars which are no longer active and therefore no longer visible. The methods developed could be applied and adapted for a broad range of scientific cases extracted from the forthcoming spectroscopic surveys.