Analog Gated Recurrent Unit Neural Network for Detecting Chewing Events
Analog Gated Recurrent Unit Neural Network for Detecting Chewing Events
复制标题
DOI：
10.1109/tbcas.2022.3218889
复制
发表时间：
2022-12-01
影响因子：
5.1
通讯作者：
Kotz，David
中科院分区：
文献类型：
--
作者：
Odame，Kofi;Nyamukuru，Maria;Kotz，David
我们提出了一种新颖的门控循环神经网络来检测一个人何时咀嚼食物。我们将神经网络实现为采用 0.18m CMOS 技术的定制模拟集成电路。该神经网络接受了 6.4 小时的数据训练，这些数据是从安装在志愿者乳突骨上的接触式麦克风收集的。当对 1.6 小时的前所未见的数据进行测试时，模拟神经网络以 24 秒的时间分辨率识别咀嚼事件。它在消耗 W 功率的同时实现了 91% 的召回率和 94% 的 F1 分数。一个基于新型模拟神经网络的用于检测整个饮食事件（例如正餐和零食）的系统会消耗估计的 W 功率。
We present a novel gated recurrent neural network to detect when a person is chewing on food. We implemented the neural network as a custom analog integrated circuit in a 0.18m CMOS technology. The neural network was trained on 6.4 hours of data collected from a contact microphone that was mounted on volunteers' mastoid bones. When tested on 1.6 hours of previously-unseen data, the analog neural network identified chewing events at a 24-second time resolution. It achieved a recall of 91% and an F1-score of 94% while consumingW of power. A system for detecting whole eating episodes—like meals and snacks—that is based on the novel analog neural network consumes an estimatedW of power.