A Bayesian binary algorithm for root mean squared-based acoustic signal segmentation

Abstract

Changepoint analysis (also known as segmentation analysis) aims at analyzing an ordered, one-dimensional vector, in order to find locations where some characteristic of the data changes. Many models and algorithms have been studied under this theme, including models for changes in mean and / or variance, changes in linear regression parameters, etc. In this work, we are interested in an algorithm for the segmentation of long duration acoustic signals; the segmentation is based on the change of the RMS power of the signal. We investigate a Bayesian model with two possible parameterizations, and propose a binary algorithm in two versions, using noninformative or informative priors. We apply our algorithm to the segmentation of annotated acoustic signals from the Alcatrazes marine preservation park in Brazil.

The e-value and the Full Bayesian Significance Test: Logical Properties and Philosophical Consequences

“…• Testing covariance structures in multivariate normal models, treating in a unified way several alternative hypotheses (often treated as special cases in the literature): [123,232]; • Testing unit root and cointegration hypotheses in time series, using plain and simple forms of prior information like flat or Jeffreys priors (no need for artificial priors): [42,51,52,234]; • Solving Bayesian classification problems and testing nested and non-nested or separate hypotheses: [8,9,[124][125][126]160]; • Analyzing systems' reliability from failure datasets: [104,143,177] • Testing dependence structures using statistical copulas: [83]; • Testing (non)-informative sampling conditions in statistical surveys: [199]; • Model selection for generalized Poisson distributions: [97,205]; • Model selection for generalized jump diffusion and Brownian motions, extremal distributions, and persistent memory processes: [6,18,122,171,172]; • Testing independence structures in contingency tables and multinomial models: [7,19,148,158]; • Software certification according to compliance conditions: [153]; • Testing market equilibrium conditions for fundamental and financial derivative asset prices: [39]; • Testing hypotheses in empirical economic studies: [41]; • Event identification in acoustic signal processing: [98][99][100]; • Testing Hardy-Weinberg equilibrium in genetics: [29,106,…”

Improving automatic processing of wildlife sound recordings

“…So far, we have discussed the analysis of a hypothetical uncorrelated Gaussian process x 0:n . Natural sound recordings, however, have a strong correlation structure, and naïvely applying changepoint detection to them could be ineffective: it was shown recently that the number of events can be over-estimated by as much as 1000-fold this way (Hubert et al, 2019).…”