Machine Learning Powered Voice Assistant

This is a machine learning powered and speech-based virtual assistant specifically designed for Raspberry Pi.

Currently supported features (more to be added soon)

• Basic conversation (greetings, questions, etc.)
• Geolocation
• Weather (temperature, humidity, etc.) based on current location
• Date/time features (day, month, year, etc.)
• Opening websites (Google, GitHub, Wikipedia, etc.) in the browser
• Creating, pausing, and deleting timers. Timers run in the background so the program does not wait for the timer to end before listening for more input.
• Repeating a certain phrase a specified number of times.
• Exiting on command

How to run code

• Download code in a directory called assistant
• Navigate to the directory in terminal/command prompt: cd assistant/
• Install the required python packages: pip install -r requirements.txt
• Get porcupine API access key (explained in next section)
• Run main.py : python main.py

How to add access_key to line 60 in main.py

• Go to https://picovoice.ai/platform/porcupine/
• Create a account and then grab your API key
• Paste your API key and DO NOT share it! 😃

How the assistant works

The following steps give a brief overview on how the assistant works:

• The program continously listens for a user to say the wake word (currently set to “Jarvis”).
• When a wake word is detected, the program uses speech recognition to determine what the user says.
• The IntentClassifier class (found in /intentclassifier/intent_classification.py) classifies the user’s intent using the Support Vector Machine (SVM) algorithm that trains on a dataset containing sample user prompts along with their intent.
• Based on the classified intent, the assistant executes the correct function. This function is found in the /assistant_functions directory. This structure makes it easy to add new functionality to the assistant.
• For more details on these functions’ implementations, see the /assistant_functions directory.