An Analysis Framework of Temporal Representation and a Model of Population Statistical Mechanisms for Auditory Stream Segregation under the ABAB Stimulus Paradigm

In complex acoustic environments, the auditory system segregates mixed inputs from multiple sound sources into perceptually distinct streams, a process known as auditory stream segregation. Although extensively studied at the neurophysiological level, existing neural metrics typically reflect time-averaged or static representations, lacking a unified framework that jointly captures temporal dynamics and segregation strength. Here, we propose a time-resolved analytical framework for ABAB alternating auditory stimulation based on a minimal temporal unit, the Single AB Unit (Sab), and introduce the Identity Fidelity Index (IFI) to quantify normalized response bias between alternating sound components. We further extend IFI to a time-resolved formulation, IFI(t), enabling continuous tracking of neural response bias over time. Using a generative model based on Laplace tuning functions and Poisson spiking processes, we show that IFI and IFI(t) robustly capture population-level response restructuring induced by frequency separation, and are consistent with classical dip ratio measures. Further, I propose a model describing how changes in frequency tuning overlap modulate observable statistical properties of neural responses (including both mean structure and fluctuation structure), thereby influencing the formation of auditory stream segregation. This provides mechanistic support for an IFI framework grounded in the measurement of instantaneous neuronal frequency selectivity. Together, this study provides a unified computational framework for describing neural temporal dynamics and auditory stream segregation, offering a quantitative perspective on the neural coding mechanisms underlying auditory scene analysis.