Abstract:
Technological developments around artificial intelligence in the last decade have led to a significant increase in applications of audio capturing and speech processing. To increase quality of sound recording and to obtain additional spatial information from sound, microphone arrays including geometrically distributed number of microphones are used widely in last decade, but dealing with microphone arrays was requiring advanced circuitry to make digital to analog conversion and also matching circuitry to ensure synchronization and high quality. Thanks to recent advances regarding MEMS Digital Microphones, we now have very small, ultra-low power and highquality microphone devices in the market. In this thesis, most recent MEMS digital microphone alternatives are studied to create a microphone array which is used in sound source localization and beam forming applications. Alternative digital data streaming interfaces by different manufacturers including I2S and TDM were tested and evaluated and then an evaluation PCB was produced to design a real-life MEMS digital microphone array device including 4 microphones in edges of a square shaped PCB. During evaluation part of the microphone arrays, time-delay estimation model was used as sound source localization since it is the one of the most common used models thanks to its robustness and fast performance for real-use cases including speech recognition and teleconference room systems. In order to calculate delay of each channels accurately, GCC-PHAT Cross-Correlation method was used and then direction arrival angle was obtained using array geometry inputs. Lastly, an enclosure-device box was also designed and 3D fabricated in order to eliminate possible effects of acoustic port outer design.
