Sound recognition actuator using Arduino and Raspberry Pi.
This project is a rough prototype and is not intended to follow along.
If you wish to make a similar system and need more details of the project, please contact me: stengine2@gmail.com .
I have a dog in the house and I love having him with me when I go to bed. However, deeper into the night, he wakes me up by skretching the door and whimpering in order to get out of the bedroom.
I wanted to solve this problem by making a device that will automatically open the door if it detects skretching and/or whimpering.
- The system is attached to a door and listens to the sound wave around the lower part of the door.
- There are two modes of operation:
- Recording Mode : User can record sounds and create labeled sound data to train a classifier.
- Test Mode (Use Mode): The system will listened to sound and open the door according to the sound classification result.
- There can be multiple labels and user can choose target labels in Test Mode. (eg: setting target labels to whimpering and skretching)
- User can train the classifier and evaluate it.
The system is consisted 2 parts, the processing and IO control, and they can be broken into smaller subsystems. They communicate through two channels. Although both channels are capable of bidirectional communication, only one directional communication was used per channel.
________ 
- How to read the diagrams:
- Data recording: (DoorOpener.py)
- This mode can be selected in the interface.
- 2 recurring inputs: Analog read data from the arduino, user input label (+data file name + serial port selection _ for once _)
- output: A .tsv file containing the data and the label
- In this mode, the program waits for serial input from the arduino and if it receives any valid data (that is, has marked start and end), it asks the user for one-lettered label. After recording datas, when the user exits the program properly, the data will be saved as a .tsv file in the ./data directory.
- Data fitting: (Classifier.py)
- input: .tsv data, file name for saving classifier
- output: .pkl of the trained classifier. This file is used in the Test Mode.
- The features (of the sound) to be collected from the raw data can be edited in the Extractor.py .
- Application (Test Mode): (DoorOpener.py)
- This mode can be selected in the interface
- input: Analog read data from the arduino (recurring), classifier file, user's target labels
- output: Actuation signal to the arduino
- Sensor (Input): Sound Sensor
- Actuation (Output): A stepper motor to pull the door knob and a servo motor to push the door open
This is an experimental project of mine, intended to learn more about Arduino and Raspberry Pi In this project, I have learned:
- Basic Linux : directories, devices, configurations etc.
- Git in Linux
- arduino-cli in Linux
- Serial communication : UART and using Tx/Rx pins
- I2C communication protocol :
- The idea of using two wires for multiple devices.
- Python SMBus library.
- Compatibility of I2C between two boards.
- ISR (Interrupt Service Routine) :
- Stepper motor cannot work in ISR...Valuable lesson.
- Using volatile variable to manage states by interrupt.
- Event handling in Arduino
- Sound Recognition : I am already a bit familiar with the ML concept.
- Gaussian Process Classification
- Iris data
- Extracting Features of sound wave signals
- Stepper motor
- Power consumption of the motors
- Measuring the current drain
- Using capacitors for stable power supply for the motors
- How consequential a servo motor's back EMF can be (Fried a Raspberry Pi and an Arduino Nano clone)