Speech Recognition in Python

4 min readMar 10, 2022

Speech Recognition is the ability of a machine or program to identify words and phrases in spoken language and convert them to a machine-readable format.

You have probably seen it on Sci-fi, and personal assistants like Siri, Cortana, and Google Assistant.

In this tutorial, you’re going to learn how to build your own python program that is capable of performing converting your sound to textual information.

Program Overview

Our Python will be capable of recording sound through a Microphone on your Computer, and then it will send the speech to google speech recognition API and return a decoded text to us.

Requirements

Installation

pip install PyAudio 
pip install SpeechRecognition

Through SpeechRecognition Library you can perform speech recognition, with support for several engines and APIs, online and offline.

In this tutorial, we are going to use Google Speech recognition API which is free for basic
uses perhaps it has a limit of requests you can send over a certain time.

Throughout this tutorial, you will learn to perform Speech Recognition using sound that is directly fed from Microphone also using
Audio Source from File

Speech Recognition from Microphone

When Performing Speech Recognition from Microphone, we need to record the audio from the microphone and then send it to google
Speech to text recognition engine and then it will give us the textual output which will print out to the Screen

Steps involved

Recording Audio from Microphone ( PyAudio)
Sending Audio to the Speech recognition engine
Printing the Recognized text to the screen

app.py

import speech_recognition as srrecognizer = sr.Recognizer()''' recording the sound '''with sr.Microphone() as source:
    print("Adjusting noise ")
    recognizer.adjust_for_ambient_noise(source, duration=1)
    print("Recording for 4 seconds")
    recorded_audio = recognizer.listen(source, timeout=4)
    print("Done recording")''' Recorgnizing the Audio '''
try:
    print("Recognizing the text")
    text = recognizer.recognize_google(
            recorded_audio, 
            language="en-US"
        )
    print("Decoded Text : {}".format(text))except Exception as ex:
    print(ex)

Speech Recognition from Audio File

When it comes to performing Speech Recognition from Audio line only one line of code is going to change instead of using a Microphone as a source of Audio, we will give a path to our Audio File we want to transcribe to text

On Demo, I have used the below sample audio

Sample Audio

Our Code looks like this

app_audio.py

import speech_recognition as srrecognizer = sr.Recognizer()''' recording the sound '''with sr.AudioFile("./sample_audio/speech.wav") as source:
    recorded_audio = recognizer.listen(source)
    print("Done recording")''' Recorgnizing the Audio '''
try:
    print("Recognizing the text")
    text = recognizer.recognize_google(
            recorded_audio, 
            language="en-US"
        )
    print("Decoded Text : {}".format(text))except Exception as ex:
    print(ex)

Output :

kalebu@kalebu-PC:~$ python3 app_audio.py 
Done recording
Recognizing the text
Decoded Text : python programming is the best of all by Jordan

Speech Recognition from long Audio Source

When you have very long audio, loading the whole audio to Memory and sending it over API can be a very slow process, to overcome that we have to split the long audio source into small chunks and then performing speech recognition on those individual chunks

We are going to use pydub to split the Long Audio Source into those small chunks

To install pydub just use pip

$ pip install pydub

Here is Sample long audio that I have used on this tutorial

Below is a sample Python code that loads the long audio, split into the segment, and then performing the Speech recognition on those individual chunks to to learn more about splitting the audio you can check out DataCamp Tutorial

long_audio.py

import os 
from pydub import AudioSegment
import speech_recognition as sr
from pydub.silence import split_on_silencerecognizer = sr.Recognizer()def load_chunks(filename):
    long_audio = AudioSegment.from_mp3(filename)
    audio_chunks = split_on_silence(
        long_audio, min_silence_len=1800,
        silence_thresh=-17
    )
    return audio_chunksfor audio_chunk in load_chunks('./sample_audio/long_audio.mp3'):
    audio_chunk.export("temp", format="wav")
    with sr.AudioFile("temp") as source:
        audio = recognizer.listen(source)
        try:
            text = recognizer.recognize_google(audio)
            print("Chunk : {}".format(text))
        except Exception as ex:
            print("Error occured")
            print(ex)print("++++++")

Output :

$ python long_audio.py
    Chunk : by the time you finish reading this tutorial you have already covered several techniques and natural then
    Chunk : learn more
    Chunk : forgetting to subscribe to be updated on upcoming tutorials
    ++++++

Hope you had a good time playing with Speech Recognition in Python

The full code for is this tutorial is available at My GITHUB PROFILE

In case of anything, comment, suggestion, difficulties, drop it in the comment box and I will get back to you ASAP.