Humming game guide: how a phone hears the pitch of your voice

What is pitch, in voice terms?

When you make a voiced sound, your vocal folds open and close over and over. Hess describes this as a quasiperiodic vibration: nearly regular, but never perfectly so. The rate of that vibration is called the fundamental frequency, usually written F0 and measured in hertz (Hz), or cycles per second. It is what we hear as the pitch of the voice.

A steady hum is a good fit for this kind of game because it is a held, voiced sound, without the stops and hisses of normal speech. That is our reading of the science rather than a claim from either source, but it matches why voice games ask you to hum rather than talk.

Why do voices sit at different pitches?

The NCVS gives typical speaking pitches as about 125 Hz for adult men and about 200 Hz for adult women, with children’s speech from about 250 to 400 Hz and baby cries around 500 Hz.

Body size is only part of the story. The NCVS explains that the main reason for the difference between adult male and female voices is the length of the part of the vocal folds that actually vibrates. On average, men’s is about 60% longer, which fully accounts for the gap.

For a voice pitch game, that means there is no single “right” note. A good game works around the range of the person playing, and a deep voice and a high voice should both be able to reach the goals.

How does a phone work out the note you hum?

Software that finds pitch is called a pitch determination algorithm. Hess explains that many of these work on short frames of sound and look for the repeating pattern inside each one. He gives 100 frames a second as a usual rate.

One of the oldest methods is autocorrelation. The program compares the sound with a slightly delayed copy of itself. When the delay matches the length of one vibration cycle, the two line up and the result shows a strong peak. The length of that cycle gives the pitch.

Hess is clear that this is harder than it sounds:

• Voices cover a wide range. For an unknown speaker, he says F0 can vary over almost four octaves, from about 50 to 800 Hz.
• Voices change quickly. The shape of the mouth and throat can shift at any moment, which changes the sound from one cycle to the next.
• Strong resonances can fool it. Hess notes that autocorrelation tends to fail when the signal has a strong resonance at the second or third harmonic. Separately, he lists octave errors, where a detector reads half or double the real pitch, among the typical mistakes.

Tips for playing a humming game

These follow from the points above:

• Hold one steady note. A detector looks for a repeating pattern, so a wobbly or breathy hum is harder to follow.
• Move between notes smoothly. Slide up or down rather than leaping, so you can see the effect before you overshoot.
• Stay in your comfortable range. Straining for very high or very low notes makes the sound less regular, and it is not kind to your voice.
• Cut background noise. Other voices, music or a television add sounds with their own pitches.
• Use headphones if the game makes sound. That stops the game’s own audio feeding back into the microphone.

Questions people ask

Where Hum Puzzle helps, and where it does not

Hum Puzzle is our Android humming game: a puzzle where you hum a note to move lifts, gates, planks and drawbridges and roll a ball to the exit. A quick humming check sets every chamber to your own range, so deep and high voices can reach the same goals. Headphones or a quiet room help it hear you. If you would rather not use the microphone, an on-screen slider plays every chamber instead. It works offline with no account, and your progress stays on your phone.

Free: Chapter 1, with 10 chambers. A one-time purchase opens Chapters 2 and 3, with 20 more.

What it is good for: a playful puzzle where your own voice is the control, with a slider for times when humming out loud is not an option.

What it will not do: it is a voice-controlled puzzle game, not singing lessons, and it does not score your singing like karaoke. How well it follows you depends on your phone’s microphone and the noise in the room.

Questions or bugs: Email us.

Sources

1. Factors Influencing Fundamental FrequencyNational Center for Voice and Speech · ncvs.org
2. Pitch and Voicing Determination of Speech with an Extension Toward Music Signals, Springer Handbook of Speech Processing (2008)W. J. Hess · doi.org
3. Pitch and Voicing Determination of Speech with an Extension Toward Music Signals, course reading copy (University of Illinois ECE 420)W. J. Hess · courses.physics.illinois.edu