Fully digital low-power implementation of an audio front-end for portable applications

Abstract

This work presents a fully digital implementation of an audio front-end for portable applications developed in a low-power perspective. Such platform acquires audio samples through a four digital Pulse Digital Modulation (PDM) microphones array and implements two-channel (left and right) beamformers to perform speech enhancement. The system also features a low-power Voice Activity Detector (VAD) for the implementation of source localization or noise reduction algorithms. Finally, the processed data can be converted into a 16-bit resolution Pulse Code Modulation (PCM) format and then transmitted through a standard Integrated Interchip Sound (I2S) interface. The front-end has been implemented onboard a Microsemi IGLOO nano Field Programmable Gate Array (FPGA), reaching a power consumption of only 2.264 mW. The low-power architecture and the implemented functionalities make the system a promising front-end for audio speech portable applications.