Test chord recognition

A tool for finding out whether chords can be recognised through the microphone at all - and for setting the thresholds.

Microphone is off

What is sounding right now

One bar per pitch class. All octaves are merged - a C is a C, whatever the position.

What the incoming sound is compared against.

The note the chord is built on.

Major or minorMajor sounds bright, minor dark.

majorminor

Similarity0 %

Good for comparing, poor as a threshold: even flat noise reaches around 50%.

Energy on chord tones0 %

How much of what is heard belongs to the chord at all. Around 25% for noise.

Not recognised

If the chord you played is not at the top, this shows what it was confused with.

They take effect immediately, without restarting the microphone.

Volume threshold (minVolume)0.010

Below this nothing is even looked for. Too high: quiet notes get missed. Too low: room noise sets it off.

Minimum similarity0.85

Good for comparing, poor as a threshold: even flat noise reaches around 50%.

Minimum energy on chord tones0.65

How much of what is heard belongs to the chord at all. Around 25% for noise.

Lowest note analysedC3

Below about C3 the semitones sit closer together than the frequency resolution - energy smears into neighbouring pitch classes there.

Current values to copy into the code: minVolume = 0.010, minSimilarity = 0.85, minChordToneEnergy = 0.65, minMidiNote = 48

It cannot tell the shape. Because all octaves are merged, E major at fret 0 sounds exactly like E major at fret 7 to the recognition.
Telling major from minor is shaky: the major third is part of every note's harmonic series, the minor third is not.
A single plucked note looks like a major chord - its harmonics supply root, fifth and major third.
Chords sharing two notes (C major, A minor, E minor) blur into one another.
The pluck itself is briefly noise across every pitch class.