/Echolocator

Primary LanguageC++

Echolocator

This project demonstrates using a computer speaker and microphone to determine distances to objects using echolocation.

Note that the distance found is a crude measure; differences in computer hardware and speaker placement make accurate distance detection impossible without prior calibration. The true distance to an object will be true_distance = a * measured_distance + b, where a is an unknown constant (close to 1) controlled by the distance between the microphone and the speaker, and b is another unknown constant controlled by the time taken between generating a signal and the speaker producing it, plus the delay between the microphone receiving a singal and the program processing it. There will also be small errors due to slight differences in the speed of sound due to, for example, air pressure.

Setup

The makefile will attempt to install dependencies in known environments:

make environment

Alternatively, you can install the dependencies yourself.

MacOS with Homebrew

brew install fftw portaudio

MacOS with MacPorts

port install fftw-3 portaudio

General

This project will build and run on unix-based systems, and possibly Windows too (all dependencies are cross-platform).

The required libraries are:

Build and Run

make
build/main

The main application window will open and the speakers will begin producing rapid clicks. After a few seconds of calibration, the display will show vertical bars representing echos received. This typically involves one strong bar (for the sound travelling directly from the speaker to the microphone), then a cluster of fuzzy bars (various reflections from the computer itself) then an emptier area. You can try moving flat objects (card, a hand, etc.) near the computer to see them appear in this area.

Theory

Sadly I can't find the original website which inspired this implementation, but I have found a page which summarises a lot of interesting research on how bats use echolocation to determine position, velocity, approximate shape, and even elevation. https://www.hscott.net/the-dsp-behind-bat-echolocation/

The general approach used here is to send out a "chirp" sound (a short tone with rapidly changing pitch), then use fourier transforms to deconvolve the returned audio. By deconvolving using the original chirp, it is possible to detect reflections with high accuracy while dismissing background noise. This allows a map of distance to reflection intensity to be produced (where time-since-chirp is used as a measure of the distance), and this map is displayed to the user.

Structure

The vague structure of this project is:

  • chirps.hpp: contains signal generators (tone, chirp and repeating chirp)
  • fourier.hpp: simple object-based wrapper around FFTW
  • audio.cpp: the main logic for the echolocation process
  • renderer.cpp: a simple wrapper around OpenGL / GLUT for displaying bitmap data
  • main.cpp: main entrypoint for the program and orchastration