How to Convert Speech to Text in Python

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

You have probably seen it on Sci-fi, and personal assistants like Siri, Cortana, and Google Assistant, and other virtual assistants that interact with through voice.

In order to understand your voice these virtual assistants need to do speech recognition.

Speech Recognition is a complex process, so I'm not going to teach you how to train a Machine Learning/Deep Learning Model to do that. Instead, I will instruct you how to do it using google speech recognition API.

As long as you have the basics of Python you can successfully complete this tutorial and build your own fully functioning speech recognition programs in Python.

Requirements

To successfully complete this tutorial, you need to have the following Python library installed on your Machine

PyAudio Library
SpeechRecognition Library

Installation

pip install PyAudio
pip install SpeechRecognition

SpeechRecognition library allows you to perform speech recognition with support for several engines and APIs, online and offline.

Below are some of the supported Engines

CMU Sphinx (works offline)
Google Speech Recognition
Google Cloud Speech API
Wit.ai
Microsoft Bing Voice Recognition
Houndify API
IBM Speech to Text
Snowboy Hotword Detection (works 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 be performing Speech Recognition using sound that is directly fed from Microphone also using audio from files.

Speech Recognition from Microphone

When Performing Speech Recognition from Microphone, we need to record the audio from the microphone. Then, we send it to Google speech to text recognition engine, which will perform the recognition and return out transcribed text

Steps involved

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

Below is a sample app.py code, it is pretty straight forward

app.py

import speech_recognition as sr

recognizer = 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 files 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

The below code is a sample script to perform speech recognition of audio in a file.

import speech_recognition as sr

recognizer = 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 it 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

To use the below link to download sample long audio

Long Sample Audio

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.

import os 
from pydub import AudioSegment
import speech_recognition as sr
from pydub.silence import split_on_silence

recognizer = 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_chunks

for 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
    ++++++

Congrats you now know how to do, can't wait to see what you're going to build with this knowledge!

To learn more Python you can visit my blog kalebujordan.com

In case of any comment, suggestion, or difficulties comment below and I will get back to you ASAP.

Previously published at https://kalebujordan.com/python-speech-recognition/