Audio Pitch Estimator for Unity

日本語

A simple C# script to estimate fundamental frequency from Unity's AudioSource component.

Just copy AudioPitchEstimator.cs to your Asset directory. This script doesn't require any external dependencies except for built-in Unity components.

Originally I made it for speech and singing, but it may be applied to some single-note instrument sounds (e.g. guitar tuner).

Demo

Movie: Pitch estimation with Unity (in Japanese)

This repository has 2 examples:

• Assets/Examples/Example1.unity
  • Estimation with audio file.
  • Includes sample audio (my singing of a public domain song).
• Assets/Examples/Example2.unity
  • Estimation with real-time audio from microphone.
  • Note: There is a delay of ~1 second because the audio signal is buffered to the AudioSource before estimation.

Usage

Please attach the AudioPitchEstimator.cs to your GameObject.

There are several parameters. You should set these parameters according to the typical characteristics of your input audio signals (e.g. the range of the voice).

• Frequency Min: The lowest frequency that can estimate.
• Frequency Max: The highest frequency that can estimate.
• Harmonics To Use: Number of overtones to use for estimation.
• Smoothing Width: Frequency bandwidth of spectral smoothing filter.
• Threshold SRH: Threshold to judge silence or not.

The boundary frequencies Frequency Min and Frequency Max should be set as nessesary and sufficient range. For example, if the fundamental frequency never exceeds 600 Hz (in your application), set Frequency Max to 600 Hz or above.

Tuning Tips

Quickly switches between two frequencies

Decreasing Frequency Max will fix the problem by eliminating unwanted high-frequency candidates.

Misidentifies silence at the tail of your speech

Increasing Threshold SRH may help. If it does not change, try tweaking Harmonics To Use and Smoothing Width.

Sometimes doesn't detect your speech

Decreasing Threshold SRH may help. If it does not change, try tweaking Harmonics To Use and Smoothing Width.

CPU load is too high

It is effective to decrease your estimation update rate (e.g. lower the PitchVisualizer.estimateRate to 8).

Example Code

Audio data can be obtained via AudioSource. If you want to use the audio from the microphones, please use Microphone.Start(), a built-in method in Unity.

AudioPitchEstimator.Estimate() takes an argument for target AudioSource and will return the estimate of fundamental frequency in the input audio signal.

void EstimatePitch()
{
    var estimator = this.GetComponent<AudioPitchEstimator>();
    var audioSource = this.GetComponent<AudioSource>();

    // Estimates fundamental frequency from AudioSource output.
    float frequency = estimator.Estimate(audioSource);

    if (float.IsNaN(frequency))
    {
        // Algorithm didn't detect fundamental frequency (e.g. silence).
    }
    else
    {
        // Algorithm detected fundamental frequency.
        // The frequency is stored in the variable `frequency` (in Hz).
    }
}

void Update()
{
    // It is NOT recommended to run the estimation every frame.
    // This will take a high computational load.
    // EstimatePitch();
}

void Start()
{
    // It is recommended to estimate at appropriate time intervals.
    // This example runs every 0.1 seconds.
    InvokeRepeating("EstimatePitch", 0, 0.1f);
}

Algorithm

This package implements a modified version of SRH (Summation of Residual Harmonics) [1], a spectrum-based pitch estimation algorithm.

Please see the original paper [1] and AudioPitchEstimator.cs for details.

License

The Unlicense. For the details, see LICENSE.

References

[1] T. Drugman and A. Alwan: "Joint Robust Voicing Detection and Pitch Estimation Based on Residual Harmonics", Interspeech'11, 2011.