(Research) STT (1) – Converting Speech to Text (Guide)

Guide to Converting Speech to Text Using Python

Those who have some experience in program development might have wondered or been intrigued by the idea of developing voice-based services when using AI speakers (or assistants). ^^

For such individuals, even those with limited development experience or lacking knowledge in natural language processing or speech recognition, there are several easy-to-follow packages that I would like to introduce.

You can utilize these tools according to your specific situation

Speech_recognition

Speech Recognition Using Python: A Guide

For those interested in voice recognition services, here's an overview of the available packages and tools you can use to convert speech to text in various languages. Below are some popular packages and APIs that support speech recognition in different languages:

Popular Speech Recognition Libraries/Packages

• apiai
• assemblyai
• google-cloud-speech
• pocketsphinx
• SpeechRecognition
• watson-developer-cloud
• wit

Speech Recognition with Language Support (Other than English)

Many libraries and APIs allow you to specify a language keyword argument to recognize speech in languages other than English. Example:

python
recognize_google(audio, language='fr-FR')  # For French

Using SpeechRecognition Library

In my case, I’m using the speech_recognition library and recognize_google() method to convert speech to text in Korean.

---

Environment Setup for Speech Recognition

1. Install SpeechRecognition Library
   Use pip to install the SpeechRecognition library:

   bash
   pip install SpeechRecognition
   

2. Install PyAudio for Microphone Access
   To access the microphone with the SpeechRecognizer, you need to install PyAudio:

   On Windows:

   bash
   pip install pyaudio
   

   On Debian Linux:

   bash
   sudo apt-get install python-pyaudio python3-pyaudio
   

---

Supported Audio File Formats

The following file formats are supported for speech recognition:

• WAV (PCM / LPCM format)
• AIFF
• AIFF-C
• FLAC (Standard FLAC format, OGG-FLAC is not supported)

---

Example Code: Converting Speech to Text (Korean)

Here’s a simple example of how to use the SpeechRecognition library to convert speech into text:

import speech_recognition as sr

recognizer = sr.Recognizer()

recognizer.energy_threshold = 300

## wav 파일 읽어오기

harvard_audio = sr.AudioFile("./TEST.wav")

# audio file을 audio source로 사용합니다

r = sr.Recognizer()

with harvard_audio as source:

    audio = r.record(source)  # 전체 audio file 읽기

# 구글 웹 음성 API로 인식하기 (하루에 제한 50회)

try:

    print("Google Speech Recognition thinks you said : " + r.recognize_google(audio, language='ko'))

except sr.UnknownValueError:

    print("Google Speech Recognition could not understand audio")

except sr.RequestError as e:

    print("Could not request results from Google Speech Recognition service; {0}".format(e))

    

"""

    recognize_bing(): Microsoft Bing Speech

    recognize_google(): Google Web Speech API

    recognize_google_cloud(): Google Cloud Speech - requires installation of the google-cloud-speech package

    recognize_houndify(): Houndify by SoundHound

    recognize_ibm(): IBM Speech to Text

    recognize_sphinx(): CMU Sphinx - requires installing PocketSphinx

    recognize_wit(): Wit.ai

"""

print(recognizer)