Finding Earthquake Victims by Voice Detection Techniques

Abstract

After an earthquake or a building collapse, victim recovery is a challenging task. Recovery methods must include the location of victims by non-visual means; human speech is one such parameter that can be used in victim and rescue operations. In this paper, we discuss the application of a voice-detection technique for the discrimination of voice and non-voice sounds, based on frequency parameters such as flux, centroid, and roll-off of audio signals. Using the cross-validation tests based on a linear discriminant analysis model, flux and centroid individually displayed the highest success rates for all categories of test sample. By combining these two parameters, the recognition rate was improved to 78% for the signals with high background noise.