Proposal of Sound Source Localisation Method using Histogram and Frequency Information on Drone Audition - Evaluation of Versatility in and Actual Outdoor Environment

Prompt search and rescue operations in disaster areas are important. Therefore,

audio-based search methods using drone-embedded microphone array that can search

even in poor lighting conditions or when victims are buried in rubble have been researched.

These methods search victims by localizing human-related sound sources, such as voices

and emergency whistles, recorded by a drone-embedded microphone array. One of the critical

issues in sound source localization with drones is large ego-noise of drones. In previous

work, we proposed a novel sound source localization method to satisfy four requirements:

high tolerance to quickly changing dynamic noise, large search range, high real-time performance,

and high versatility. This method estimates and separates the ego-noise and target

sound components of the spatial spectrum calculated by multiple signal classification using

histogram and frequency information at the current time. The evaluation results of numerical

simulations confirmed that the proposed method satisfies four requirements simultaneously.

However, the evaluation of the versatility was not sufficient because only two drones were

used in the numerical simulations. In this report, the versatility of the proposed method

is further evaluated. The performance of the proposed method was evaluated using three

different drones in an actual outdoor environment. The results confirm that the proposed

method has high versatility in practical use. It is also found that improving the tolerance to

wind noise is an issue to be addressed in future work.