Could run on a Pi Zero W with a ReSpeaker 2-Mics Pi HAT to capture 48kHz 2 channel audio and process it with the McLeod's WSNAC pitch detection algorithm, thus implementing a low latency tuner and intonation monitor.

N.B. it should run on your Linux too.

Tha audio samples are captured using ALSA and than processed with the SNAC algorithm in a tiny C backend.

Using the rawrtc implementation of the WebRTC DataChannel, the pitch data could then be sent to a web application for visualization.

The WSNAC implementation uses currently multiple FFTs of size 2048 for processing an audio frame captured every 1024/48kHz seconds. Using the kiss_fft on the Zero Pi needs less than 40% CPU, and with GPU_FFT it comes down to 18% CPU.

First make sure you have CMake 3.2+ and the ALSA dev libs:

sudo apt-get install cmake libasound2-dev

To build and install (locally with a prefix under ./bulid/prefix):

export PKG_CONFIG_PATH=${PWD}/build/prefix/lib/pkgconfig
cd build
cmake -DCMAKE_INSTALL_PREFIX=${PWD}/prefix ..
make install

N.B. You might want to

sudo apt-get install libssl-dev

before doing ./make-dependencies.sh on the Zero Pi, as compliling openssl from sources would take quite some time.

To check if it works:

export LD_LIBRARY_PATH=${PWD}/prefix/lib/:${LD_LIBRARY_PATH}
./build/prefix/bin/zerotune-test hw:1,0

N.B. When using GPU_FFT you might have to run it with sudo.

Raspberry Pi Zero W

ReSpeaker 2-Mics Pi HAT

McLeod, Philip - Fast, Accurate Pitch Detection Tools for Music Analysis, 2008.

Kiss FFT

Accelerating Fourier Transforms using the GPU

hello_fft at RaspberryPi Firmware.