BigEar: Ubiquitous Wireless Low-Budget Speech Capturing Interface

• 作者: Andrea Masciadri Fabio Salice Sara Comai Stefano Gorla
• 来源期刊: 电脑和通信(英文)
• 学科: 工学
• 关键词: WIRELESS Sensor Networks SPEECH CAPTURE Degraded SPEECH Recognition UBIQUITOUS Systems LOW Cost ARCHITECTURES
• 年，卷（期）: 2017,（4）
• 页码范围: 60-83
• 页数: 24页
• 分类号: TP39

摘要

This article presents BigEar, a wireless low-cost speech capturing interface that aims to realize unobtrusive and transparent context-aware vocal interaction for home automation. The speech recognition process implemented in BigEar system considers noise sources including possible holes in the reconstructed audio stream and tries to overcome them by means of inexactness toleration mechanisms to improve intelligibility of the reconstructed signal. Key contribution of this work is the use of extremely low cost devices to realize a modular flexible and real-time wireless sensor network. On-field implementation and experiments show that the proposed solution can perform real-time speech reconstruction, while listening tests confirm the intelligibility of the reconstructed signal.