Audio processing using Pydub and Google speechRecognition API

Audio files are a widespread means of transferring information. So, let's see how to break down audio files (.wav files) into smaller chunks, and to recognize the content in them and store it to a text file. To know more about audio files and their formats, refer Audio_formats. Need to break down an audio file? When we do any processing on audio files, it takes a lot of time. Here, processing can mean anything. For example, we may want to increase or decrease the frequency of the audio, or as done in this article, recognize the content in the audio file. By breaking it down into small audio files called chunks, we can ensure that the processing happens fast. Required Installations:

pip3 install pydub
pip3 install audioread
pip3 install SpeechRecognition

There are majorly two steps in the program. Step #1: It deals with slicing the audio files into small chunks of a constant interval. The slicing can be done with, or without overlap. Overlap means that the next chunk created will start from a constant time backward, so that during the slicing if any audio/word gets cut, it can be covered by this overlap. For example, if the audio file is 22 seconds, and the overlap is 1.5 seconds, the timing of these chunks will be:

 chunk1 : 0 - 5 seconds
 chunk2 : 3.5 - 8.5 seconds
 chunk3 : 7 - 12 seconds
 chunk4 : 10.5 - 15.5 seconds
 chunk5 : 14 - 19.5 seconds
 chunk6 : 18 - 22 seconds

We can ignore this overlap by setting the overlap to 0. Step #2: It deals with working with the sliced audio file to do whatever the user requires. Here, for demonstration purposes, the chunks have been passed through the Google Speech recognition module, and the text has been written to a separate file. To understand how to use the Google Speech Recognition module to recognize the audio from a microphone, refer this. In this article, we will be using the sliced audio files to recognize the content. Step #2 is done in a loop inside Step #1. As soon as the audio file is sliced into the chunk, the chunk is recognized. This process continues till the end of the audio file. Example:

Input : Geek.wav


Output : 
Screenshot of cmd running the code:

Text File: recognized

  Below is the implementation: 

Output: 👁 Image
recognized.txt - 👁 Image
As we can see in the above screenshot, all these chunks are stored in the local system. We have now successfully sliced the audio file with an overlap and recognized the content from the chunks. Advantages of this method:

• The interval can be set to any length depending on how long we need the chunks to be.
• Overlap ensures that no data is lost even if any word is said precisely at the end of the interval.
• The chunks can all be stored in different audio files and used later if need be.
• Any processing which can be done on an audio file can be done in these chunks as well, as they are just audio files.

Disadvantages of this method:

• Using Google Speech Recognition requires an active internet connection.
• After the overlap, some text processing should be done to remove the duplicate words recognized.
• The accuracy of Google Speech Recognition varies on a lot of factors like background noise, speaker's accent etc.