谷歌语音识别使用多少带宽？ [英] How much bandwidth does google speech recognition use?

问题描述

我使用 https://github.com/TalAter/annyang 将语音翻译为文字。

在调用了 annyang.start（）; 之后，我发现麦克风连接到谷歌的语音识别引擎并且只会在调用 annyang.abort（）时停止流数据。

必须知道我可能会让麦克风进入被动语音检测状态，或者在命令发出后需要关闭麦克风以保留数据和带宽。

我无法找到有关谷歌语音识别使用多少数据的统计数据。

有没有关于这个话题的任何文件，我错过了或者我怎么可以自己来衡量这个？

解决方案

不知道这是否是标准的，但这是我的发现：

我看了一下网络活动。
$ b

在启动语音识别服务后，我的出站带宽从每秒几个字节（0〜60 bps）变为每秒约18KB〜20KB，每秒大约+ - 1KB〜1.5 KB即时消息。

无论我做了什么噪音，这个速度似乎都相当稳定。

+ - 20KB per第二次60 * 60 = 72000KB。

结论：

语音识别将会如果服务处于活动状态，则每小时消费72美元。

I am using https://github.com/TalAter/annyang to translate speech into text.

After annyang.start(); has been called, I discovered that the microphone connects to google's speech recognition engine and will only stop streaming data once annyang.abort() is called.

It is imperative to know whether I could leave the mic on for passive voice detection or if it is necessary to turn the mic off after a command is given in order to preserve data and bandwidth.

I could not find any statistics regarding how much data google's speech recognition uses.

Is there any documentation on this topic that I missed or how could I measure this myself?

解决方案

Not sure if this is standard but here is my findings:

I had a look at the network activity.

After starting the speech recognition service my outgoing bandwidth went from a couple of bytes per second (0 ~ 60 bps) to about 18KB ~ 20KB per second outgoing and about +- 1KB ~ 1.5 KB per second incoming.

This rate seemed to be pretty steady regardless of what noises I made.

+- 20KB per second times 60*60 = 72000KB.

Conclusion:

Voice recognition will eat about 72meg per hour if the service is left active.

这篇关于谷歌语音识别使用多少带宽？的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！