<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
	<channel>
		<title>唤醒词 on 编程心语</title>
		<link>https://www.ithome.me/tags/%E5%94%A4%E9%86%92%E8%AF%8D/</link>
		<description>Recent content in 唤醒词 on 编程心语</description>
		<generator>Hugo</generator>
		<language>zh-CN</language>
		
		
		
		
			<lastBuildDate>Mon, 10 Aug 2026 03:03:44 +0800</lastBuildDate>
		
			<atom:link href="https://www.ithome.me/tags/%E5%94%A4%E9%86%92%E8%AF%8D/index.xml" rel="self" type="application/rss+xml" />
			<item>
				<title>端侧唤醒词检测实战：给语音助手装上离线守门员</title>
				<link>https://www.ithome.me/post/2026/08/10/wake-word-kws-speech/</link>
				<pubDate>Mon, 10 Aug 2026 08:00:00 +0800</pubDate>
				<guid>https://www.ithome.me/post/2026/08/10/wake-word-kws-speech/</guid>
				<description>&lt;h2 id=&#34;为什么语音助手需要一个守门员&#34;&gt;为什么语音助手需要一个守门员&lt;/h2&gt;&#xA;&lt;p&gt;智能音箱、车载助手、会议耳机——只要涉及语音交互，第一个问题都是：&lt;strong&gt;怎么知道用户开始说话了？&lt;/strong&gt;&lt;/p&gt;&#xA;&lt;p&gt;最笨的方案是全时跑 ASR。但上一篇里那个 24MB 的 int8 识别模型，全时转写依然会吃掉可观的 CPU 和电量。更聪明的做法是三级接力：&lt;/p&gt;</description>
			</item>
	</channel>
</rss>
