Method for automatically capturing and tracking speaker

Abstract

The invention relates to a method for automatically capturing and tracking a speaker, which comprises the following steps that: (1) a microphone array collects an external sound signal and sends the external sound signal to a central processing unit, and the central processing unit analyzes whether an effective sound input exists; (2) it is determined whether the sound signal is a human voice signal or not; (3) the orientation of the sound is analyzed by using a sound source localization algorithm, a rotation angle of a camera module is calculated according to the orientation of the sound, and a control instruction is sent to a rotary table according to the rotation angle; (4) the rotary table adjusts the position of the camera module according to the control instruction, the camera module captures video data and sends the video data to the central processing unit in the position adjusting process, and the central processing unit analyzes whether a face is captured in a captured picture in real time by using a face recognition algorithm; and (5) the central processing unit determined whether the captured face image is optimal or not in real time. According to the invention, the lens of the camera can always quickly capture and track the current speaker, and capture and tracking are accurate and high in precision.

Description

technical field [0001] The invention relates to the technical field of fusion of sound and image information, in particular to a method for automatically capturing and tracking a speaker. Background technique [0002] Currently, in common video call systems, the position and direction of the camera are fixed. In order to achieve an ideal video call effect, the two or more parties in the call must face the camera within a specified range to facilitate the camera to capture images. However, in actual use, there are often some application scenarios that cannot meet this requirement, such as: [0003] (1) One party of the video call cannot master the knowledge of the video call. For example, underage children and left-behind elderly people cannot master the essentials of video calling due to their limited knowledge. Children are restless by nature and cannot stay fixed in one fixed position. However, video calls for these two types of groups are often more urgent needs. [0...

AI Technical Summary

Problems solved by technology

[0004] (2) The party on the call itself is in a state of motion and cannot be fixed at a certain position

The solution first requires additional manpower to carry out this operation

Secondly, the speed of manual adjustment is slow and the accuracy is poor. It is difficult to quickly capture the speaker and keep up with the rhythm of speaker switching.

Images