简体繁体 English

在android有没有办法比较两个音频文件之间的相似性？

[英]In android is there a way to compare two audio files for similarity between them?

原文 2015-11-17 11:13:54 5 1 android

I am currently working on a game that requires me to record an audio file,which is made up of a text read by a person. 我目前正在制作一款游戏，要求我录制一个音频文件，该文件由一个人阅读的文本组成。 This particular audio file needs to be compared with another (default) audio file, for the recording of which same text as before was used (while creating the game). 需要将此特定音频文件与另一个（默认）音频文件进行比较，以记录与之前相同的文本（在创建游戏时）。 Now I need to know how similar these two files are. 现在我需要知道这两个文件有多相似。 Is there a way to do that? 有没有办法做到这一点？ A function that returns me a similarity index? 一个函数，它返回一个相似性指数？ Even if I have to integrate an external library, its fine. 即使我必须集成一个外部库，它的罚款。

I have read online about using FFT and correlation methods to do that, but is there a library that has all those techniques already built in it? 我在网上已经阅读了使用FFT和相关方法来做到这一点，但是有没有一个库已经内置了所有这些技术？

1 个解决方案

Yes, te base of this library should be FFT (fast Fourier transform) for taking all mixing of every single frequence of your files. 是的，这个库的te base应该是FFT （快速傅立叶变换），用于对文件的每个频率进行所有混合。 It means, by a short way, 'transform audio into frequences (number) by time axis'. 这意味着，通过一种简短的方式，“按时间轴将音频转换为频率（数量）”。 But when someone is reading, there can be a diferent ways to make it with different frequences. 但是当有人在阅读时，可能会有不同的方法来制作不同的频率。 The differentiation of voice is not frecuences, is modulation . 语音的区别不是自由，而是调制。 Then, you will need harder FFT library. 然后，您将需要更难的FFT库。 Maybe with 'Speech-Recognition' (witch includes FFT modulation) you will be able to make it possible. 也许通过“语音识别”（包括FFT调制），您将能够实现这一目标。