Sound source direction estimation method and device based on time-frequency masking and deep neural network

Abstract

The disclosure discloses a sound source orientation estimation method, device, electronic equipment, and storage medium based on time-frequency masking and deep neural networks, and belongs to the field of computer technology. The method includes: acquiring a multi-channel sound signal; performing framing, windowing and Fourier transform on each channel sound signal in the multi-channel sound signal to form a short-time Fourier transform of the multi-channel sound signal Spectrum; The short-time Fourier spectrum is iteratively calculated by the pre-trained neural network model, and the ratio film corresponding to the target signal in the multi-channel sound signal is calculated, and a plurality of ratio films are fused to form a single ratio film; The single ratio membrane performs masking and weighting on the multi-channel sound signal to determine the orientation of the target sound source. The above sound source direction estimation method and device based on time-frequency masking and deep neural network can have strong robustness in environments with low signal-to-noise ratio and strong reverberation, and improve the accuracy and stability of target sound source direction estimation.

Description

technical field [0001] The present disclosure relates to the field of computer application technology, and in particular to a sound source direction estimation method, device, electronic equipment, and storage medium based on time-frequency masking and deep neural networks. Background technique [0002] Sound source localization in noisy environments has many real-life applications, such as human-computer interaction, robotics, and beamforming. Traditionally, GCC-PHAT (Generalized Cross Correlation Phase Transform, generalized cross-correlation-phase transformation method), SRP-PHAT (Steered Response Power Phase Transform, phase transformation weighted controllable response power method) or MUSIC (Multiple Signal Classification, multiple Signal classification) and other sound source localization algorithms are the most common. However, these algorithms can only localize the loudest signal sources in the environment, which may not be the target speaker at all. For example, ...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the technical problem of poor robustness of orientation estimation, the present disclosure provides a sound source direction estimation method, device, electronic equipment, and storage medium based on time-frequency masking and deep neural network

Images