Master Python Speech Recognition: From Basics to Real-World Audio Transcription

Overview of Speech Recognition Working Principles

Speech recognition originated in the early 1950s at Bell Labs, initially handling a single speaker and a tiny vocabulary. Modern systems can recognize multiple speakers and support large multilingual vocabularies.

The process starts with a microphone converting sound into an electrical signal, which is digitized via an ADC. The digital signal is then processed by models to transcribe audio into text.

Most contemporary systems rely on Hidden Markov Models (HMM) that treat short‑time frames (e.g., 10 ms) as quasi‑static processes. Neural networks often preprocess the signal—performing feature extraction, dimensionality reduction, or voice activity detection—before HMM decoding.

Choosing a Python Speech‑Recognition Package

Several packages are available on PyPI, including:

apiai

google-cloud-speech

pocketsphinx

SpeechRecognition

watson-developer-cloud

wit

Some, like wit and apiai , add natural‑language‑understanding capabilities. Google Cloud Speech focuses on speech‑to‑text conversion. The SpeechRecognition library stands out for its ease of use.

Installing SpeechRecognition

$ pip install SpeechRecognition

After installation, verify the version in a Python interpreter:

>>> import speech_recognition as sr
>>> sr.__version__
'3.8.1'

SpeechRecognition works with Python 2.6, 2.7, and 3.3+. This tutorial assumes Python 3.3+.

Using Audio Files

Download sample audio files (e.g., GitHub repository ) and place them in the script directory.

Supported file types are WAV (PCM/LPCM), AIFF, AIFF‑C, and FLAC (native FLAC only). On Linux/macOS/Windows, FLAC support is required; otherwise install the FLAC encoder. $ pip install pyaudio Load an audio file with sr.AudioFile and record its contents:

>>> harvard = sr.AudioFile('harvard.wav')
>>> with harvard as source:
...     audio = r.record(source)
>>> type(audio)
<class 'speech_recognition.AudioData'>
>>> r.recognize_google(audio)
'the stale smell of old beer lingers it takes heat to bring out the odor a cold dip restores health and zest a salt pickle taste fine with ham tacos al Pastore are my favorite a zestful food is the hot cross bun'

Extracting Audio Segments with Offset and Duration

Use the duration argument to limit recording time, or offset to start later in the file:

>>> with harvard as source:
...     audio = r.record(source, duration=4)
...>>> r.recognize_google(audio)
'the stale smell of old beer lingers'

Combining offset and duration enables precise slicing, though inaccurate values can degrade transcription quality.

Impact of Noise on Speech Recognition

Background noise reduces accuracy. For noisy files (e.g., jackhammer.wav), apply adjust_for_ambient_noise() before recording:

>>> with jackhammer as source:
...     r.adjust_for_ambient_noise(source)
...     audio = r.record(source)
...>>> r.recognize_google(audio)
'still smell of old beer vendors'

Adjust the analysis window with the duration parameter (minimum 0.5 s recommended).

To retrieve all possible transcriptions, set show_all=True:

>>> r.recognize_google(audio, show_all=True)
{'alternative': [{'transcript': 'the snail smell like old Beer Mongers'}, ...], 'final': True}

Using a Microphone

Install PyAudio to access the microphone. Installation varies by OS:

Debian/Ubuntu: $ sudo apt-get install python-pyaudio python3-pyaudio macOS: $ brew install portaudio then $ pip install pyaudio Windows: $ pip install pyaudio Test the installation with: $ python -m speech_recognition Capture microphone input using a context manager:

>>> import speech_recognition as sr
>>> r = sr.Recognizer()
>>> mic = sr.Microphone()
>>> with mic as source:
...     r.adjust_for_ambient_noise(source)
...     audio = r.listen(source)
...>>> r.recognize_google(audio)
'hello'

If the audio cannot be matched, speech_recognition.UnknownValueError is raised; wrap calls in try/except blocks.

Conclusion

The tutorial demonstrates English speech recognition with the SpeechRecognition library, but the same methods work for other languages by passing the appropriate language code to the recognizer functions.

Author: David Amos Original article: https://realpython.com/python-speech-recognition/