Independent Reproduction and Experimental Validation of the ESS-X Audio-Based Fracture Localization Framework

This repository contains an independent experimental reproduction of the ESS-X audio-based fracture localization framework using newly acquired acoustic recordings. The study reproduces the complete ESS-X analysis workflow originally described in the Harvard Dataverse repository: Doijiri, O. (2026). Replication Data for: Audible Sound Processing Algorithm and Dataset Request Specification for Non-Invasive Bone Fracture Estimation. Harvard Dataverse, V1. https://doi.org/10.7910/DVN/A0AVBK The original analysis pipeline was implemented without modification, including: audio preprocessing transfer-function estimation spectral coherence analysis phase-response analysis cross-correlation analysis phase-locking value (PLV) Kullback–Leibler divergence spectral feature extraction Robust Z-score normalization weighted damage scoring Softmax probability estimation Independent acoustic recordings were analyzed using the published framework to evaluate methodological reproducibility. The reproduced analysis identified Edge 4–3 as the most probable abnormal structural connection with an estimated probability of 100.0%, demonstrating successful reproduction of the published computational workflow under an independent experimental setting. This repository is intended as an open reproducibility resource for researchers interested in acoustic structural analysis, computational signal processing, and non-invasive fracture localization.