“…Determining what constitutes noise in recordings is non-trivial and impacts what type of noise reduction algorithm can and should be used. In a systematic review of noise reduction methods in bio-acoustics, Xie et al ( 2020 ) outline six classes of noise reduction algorithms used for bio-acoustics: (1) Optimal FIR filter (e.g., Kim et al, 2000 ), (2) spectral subtraction (e.g., Boll, 1979 ; Kiapuchinski et al, 2012 ; Sainburg et al, 2020b ), (3) minimum-mean square error short-time spectral amplitude estimator (MMSE-STSA; e.g., Ephraim and Malah, 1984 ; Alonso et al, 2017 ; Brown et al, 2017 ) (4) wavelet based denoising (e.g., Ren et al, 2008 ; Priyadarshani et al, 2016 ) (5) image processing based noise reduction, and (6) deep learning based noised reduction. These noise reduction algorithms can be broadly divided into two categories: stationary and non-stationary noise reduction ( Figure 1A ).…”