An Open Audio Processing Platform with Zync FPGA

This paper presents an open audio processing platform on Zync7000 Field-Programmable Gate Array (FPGA), capable of collecting analog frequency signals through a microphone, and pushing out a data set of frequencies and amplitudes to a UART interface. The validity of this platform has been proved by a design on a real-time automatic music transcription (AMT) system, mainly containing three algorithms run on the Zync processor including Discrete Fourier Transform (DFT), vector building with Pitch Class Profile (PCP), and pattern matching. Experimental results show that the accuracy of the chord detection can reach 90% while playing open chords, and the latency of analog sampling is around 250 ms. We make this platform publicly available to encourage future research in audio processing such as speech recognition and natural language processing, and to serve educational studies on digital system and embedded system implementations with hardware-software co-design FPGAs.