美国陆军资助算法研究以解码脑信号

米拉碧玉

2020年11月16日

该算法是最终建立机器脑接口的工作的一部分。

美国陆军正在资助研究一种新的机器学习算法，以便可以成功地确定哪些特定行为（例如步行和呼吸）属于哪个特定的大脑信号，并且它有可能帮助军队维持更加随时可用的部队。

在任何给定时间，人们都会执行许多任务。与这些任务相关的所有大脑和行为信号混合在一起，形成一个复杂的网络。到目前为止，该网络一直难以解开和翻译。

但是，根据2020年11月12日的新闻稿，由美国陆军资助的研究人员开发了一种可以对这些信号进行建模和解码的机器学习算法。这项使用标准大脑数据集进行分析的研究最近发表在《自然神经科学》杂志上。

“我们的算法首次可以分离与特定行为有关的大脑信号中的动态模式，并且在解码这些行为方面要好得多，”南加州大学工程学教授Maryam Shanechi博士领导了这项研究，在一份声明中说。

美国陆军作战能力发展司令部陆军研究实验室一部分的陆军研究办公室项目经理Hamid Krim博士对Nextgov Shanechi说，她的团队使用该算法将所谓的与行为相关的脑信号与与行为无关的脑信号分开。

克林姆说：“这提供了一种可靠的方法，例如可以可靠地衡量个人，士兵的精神负担。”克里姆说，如果该算法检测到表明士兵处于压力或超负荷状态的行为，则机器可以在士兵甚至无法识别自己的疲劳之前对其发出警报。他补充说，提高自我意识是陆军对这项研究兴趣的关键。

该研究是建立机器脑接口的工作的一部分。克里姆说，最终，这项研究可能会促进技术的发展，该技术不仅可以解释大脑的信号，还可以将信号发送回去，以帮助人们针对某些行为采取自动纠正措施。谈到这项技术的潜力，想象力是唯一的限制。另一个未来派的应用程序可以使士兵之间相互交流而无需张开嘴巴。

克里姆说：“如果你在剧院里，不会说话，你甚至都不会低声说话，但你仍然可以交流。” “如果您可以与您的机器对话，并且该机器与另一台机器对话，并且该机器与另一名士兵对话，那么您基本上就可以建立完整的链接，而无需说一句话。”

Army-Funded Algorithm Decodes Brain Signals

A new machine-learning algorithm can successfully determine which specific behaviors—like walking and breathing—belong to which specific brain signal, and it has the potential to help the military maintain a more ready force.

At any given time, people perform a myriad of tasks. All of the brain and behavioral signals associated with these tasks mix together to form a complicated web. Until now, this web has been difficult to untangle and translate.

But researchers funded by the U.S. Army developed a machine-learning algorithm that can model and decode these signals, according to a Nov. 12 press release. The research, which used standard brain datasets for analysis, was recently published in the journal Nature Neuroscience.

“Our algorithm can, for the first time, dissociate the dynamic patterns in brain signals that relate to specific behaviors and is much better at decoding these behaviors,” Dr. Maryam Shanechi, the engineering professor at the University of Southern California who led the research, said in a statement.

Dr. Hamid Krim, a program manager at the Army Research Office, part of the U.S. Army Combat Capabilities DevelopmentCommand’s Army Research Laboratory,told Nextgov Shanechi and her team used the algorithm to separate what they call behaviorally relevant brain signals from behaviorally irrelevant brain signals.

“This presents a potential way of reliably measuring, for instance, the mental overload of an individual, of a soldier,” Krim said.

If the algorithm detects behavior indicating a soldier is stressed or overloaded, then a machine could alert that soldier before they are even able to recognize their own fatigue, Krim said. Improving self-awareness is central to the Army’s interest in this research, he added.

The research is part of an effort to establish a machine-brain interface. Eventually, Krim said, this research may contribute to the development of technology that can not only interpret signals from the brain but also send signals back to help individuals take automatic corrective action for certain behaviors, he added.

Imagination is the only limit when it comes to the potential of this technology, Krim said. Another futuristic application could enable soldiers to communicate with each other without ever opening their mouths.

“If you’re in the theater, and you can’t talk, you can’t even whisper, but you can still communicate,” Krim said. “If you can talk to your machine, and the machine talks to the other machine, and the machine talks to the other soldier, you have basically a full link without ever uttering a word.”

——END——