Design and implementation of three-dimensional localization and separation system in multi-sound source scenarios

　　Project background and feasibility analysis

　　Project name, main content and current progress

　　Main content of the project: The research on sound separation has important theoretical and practical value in sound communication, acoustic target detection, etc.; sound source separation technology has special applications in machine hearing, security monitoring, military and other fields. At present, single sound source positioning and enhancement have been applied in noisy backgrounds; but the positioning and separation in multi-sound source scenarios are rarely applied due to the complexity of algorithms and hardware. This project builds microphone array signal acquisition hardware and implements FPGA sound separation algorithm to complete the three-dimensional positioning and separation of two or more sound sources, and uses the parallelism of FPGA to achieve real-time goals. The difficulty of the project lies in making signal acquisition and processing boards, screening effective and easy-to-implement digital signal algorithms and implementing them.

　　II. Project Implementation Plan

　　The hardware system of the multi-sound source localization and separation system is divided into three parts: sound signal analog front end, data acquisition motherboard, and digital signal processing core, as shown in Figure 1.

　　Figure 1 System hardware block diagram

　　The front-end amplifier uses the high-frequency low-noise amplifier NE5532, and the analog-to-digital converter uses the AD73360 that can synchronously sample 6 channels at 64KHz. For the convenience of expansion and debugging, the data acquisition board and the signal processing board are connected by USB, and the data is more than 3MByte. Among them, the cache is used to temporarily store the intermediate results of the data, the SD card is used to store the audio data for a long time, and the human-computer interaction devices include buttons, LCD, etc.

　　Digital signal processing is the core of this project, which is divided into functional modules, including front-end correction, correlation analysis, coordinate solution, sound source separation, etc., as shown in Figure 2.

　　Figure 2 System functional module block diagram

　　The data signal processing algorithm is the focus and difficulty of this project, which requires repeated experiments and demonstrations to be determined.