Multipurpose microphone array using Raspberry Pi and MATRIX Creator

• Main Author: Toh, Darren Ting Yu
• Other Authors: Chng Eng Siong
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Computer science and engineering::Computer applications::Computer-aided engineering; Engineering::Computer science and engineering::Hardware::Input/output and data communications
• Online Access: https://hdl.handle.net/10356/144643
• Institution: Nanyang Technological University

Microphone arrays have become increasingly popular due to their usefulness in a variety of speech related applications. A lot of studies have been conducted on microphone arrays, with commercial products such as Amazon Echo enjoying widespread usage. However, existing commercial microphone array systems are either not easily customizable, difficult to deploy or too costly to be used in microphone array research and development. Thus, this project aims to develop a microphone array system that is both overcomes these issues while still being robust enough to handle heavy usage. This project uses MATRIX Creator alongside Raspberry Pi as the hardware of choice due to their ease of deployment, customization, and their relatively small size. A mobile application was developed in tandem with this system to provide a wireless architecture. WiFi tethering is used to form an ad-hoc network of any number of microphone array devices that can operate synchronously for real world applications and provides a method to upload saved data to a cloud server via cellular network. To ensure that the system is robust enough to handle heavy usage, wake word detection is used to conserve computational power and power consumption. The system also introduces Voice Activity Detection (VAD) to only save voiced data and discard the remaining data to conserve memory storage. The battery-life of the system can last up to 100 hours in passive-listening mode and up to 30 hours of active recording usage which is sufficient for practical applications. Audio samples were tested to ensure the best sample accuracy between devices and the results showed a sample deviation of around 30 samples, showing a large improvement over the non-synchronized version which has a sample deviation of around 500 samples. Overall, the developed system has met the objectives of being easy to customize and deploy while still meeting real-world use case requirements. Therefore, this project could potentially bridge the gap between developers and end-users while helping to facilitate future research and development on microphone arrays.