Scalable real-time sound source localization method based on TDOA

Sound source localization remains a critical task for increasing sufficiently the location accuracy and typically decreasing the computational complexity of a real-time passive or active target tracking by multiple-microphone in topology-based concurrent sensor arrays for far- or near-filed domains. This paper properly presents a passive real-time localization method scientifically based on the time difference of arrival signal in reliably estimating the arrival angles for meaningfully improving the location accuracy and typically decreasing the prediction time of a stationary source in a three-dimensional (3D) model for an outdoor environment in free-field conditions. The proposed scalable distributed method significantly increases the localization accuracy by integrating local direction-finding information of two concurrent parallel microphone arrays and naturally deriving a specific piece of global location-finding information to sufficiently reduce the predictions' run-time. The designing approach accurately represents a two-step process to properly implement a suitable decomposition method and gently apply an interpolation technique in the near-filed ideal domain. The primary goal of our innovative design realistically is to seemingly indicate the key role of hardware architecture and popularly use suitable techniques in complexity reduction and accuracy improvement of the intelligent model for future scalable localization systems to progressively increase outstanding performance. The empirical experiments using different real and simulated datasets show the modern TDOA-SSL method ordinarily has a low-time 360 msec in 3D accurately estimating the prime location with an average error of 12.07 cm in the near-field and 296 cm in the far-field. In key addition, the proposed method invariably is robust to ambient noise and sound reflection in the near- and far-field. However, the average accuracy of the proposed system efficiently is 99.43% with an error factor of 0.19% for the near-field and 99.70% with an error factor of 0.016% for the far-field for the used range of 100 m and 1 km, respectively. Performance evaluation of the proposed method sufficiently shows the real-time prediction and the reasonable accuracy of the specific prediction positively enhances by carefully applying the geometric architectural specification.