使用javascript语音识别质量（语音到文本）的差异 [英] differences in speech recognition quality (speech to text) with javascript

问题描述

嗨 - 我一直在使用随SDK提供的javascript示例运行一些测试。我发现该示例有时可以正常工作，但大多数情况下它并没有正确地将我的语音翻译成文本。

Hi - I have been running some tests using the javascript sample provided with the SDK. I find that the sample works sometimes, but most often it does not correctly translate my speech into text.

当我运行此页面上的示例时，https：//www.microsoft.com / cognitive-services / en-us / speech-api，翻译要好得多。 

When I run the sample found on this page https://www.microsoft.com/cognitive-services/en-us/speech-api, the translation is much better. 

我使用相同的浏览器和相同的麦克风，我说的是相同的短语，但在微软页面上找到的官方样本似乎比sdk中提供的样本更好

I am using the same browser and the same microphone and I am speaking the same phrases, but the official sample found on the microsoft page just seems to be way better than the sample provided in the sdk

还有其他人有这种经历吗？你找到了根本原因吗？

Has anyone else had this experience? Did you find a root cause?

我想知道不同的牛津键是否允许来电者访问不同的识别引擎。

I wonder if the different Oxford keys give the caller access to different recognition engines.

微软的人，你能来吗？建议和/或帮助改善语音识别使用javascript sameple？

Microsoft folks, can you advise and/or help with improving speech recognition using the javascript sameple?

谢谢！

推荐答案

当我将一个wav文件发送到Project Oxford，我得到了一致的可重复结果。语音质量很好，类似于SoundHound，不如谷歌。

When I send a wav file to Project Oxford, I get consistent repeatable results. The speech to text quality is good, similar to SoundHound, not as good as Google.

这篇关于使用javascript语音识别质量（语音到文本）的差异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！