Class PitchShift<T>

Namespace

AiDotNet.Augmentation.Audio

Assembly

AiDotNet.dll

Shifts the pitch of audio without changing tempo using WSOLA (Waveform Similarity Overlap-Add).

public class PitchShift<T> : AudioAugmenterBase<T>, IAugmentation<T, Tensor<T>>

Type Parameters

T

The numeric type for calculations.

Inheritance

object

AugmentationBase<T, Tensor<T>>

AudioAugmenterBase<T>

PitchShift<T>

Implements

IAugmentation<T, Tensor<T>>

Inherited Members

AudioAugmenterBase<T>.SampleRate

AudioAugmenterBase<T>.GetDuration(Tensor<T>)

AudioAugmenterBase<T>.GetSampleCount(Tensor<T>)

AugmentationBase<T, Tensor<T>>.NumOps

AugmentationBase<T, Tensor<T>>.OnAugmentationApplied

AugmentationBase<T, Tensor<T>>.Name

AugmentationBase<T, Tensor<T>>.Probability

AugmentationBase<T, Tensor<T>>.IsTrainingOnly

AugmentationBase<T, Tensor<T>>.IsEnabled

AugmentationBase<T, Tensor<T>>.Apply(Tensor<T>, AugmentationContext<T>)

AugmentationBase<T, Tensor<T>>.ApplyAugmentation(Tensor<T>, AugmentationContext<T>)

AugmentationBase<T, Tensor<T>>.GetParameters()

AugmentationBase<T, Tensor<T>>.RaiseAugmentationApplied(AugmentationContext<T>, bool)

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

For Beginners: Pitch shifting makes audio sound higher or lower, like the difference between a high and low voice, while keeping the same duration.

Algorithm: Uses WSOLA (Waveform Similarity Overlap-Add) for time-stretching followed by resampling. This preserves audio quality better than simple resampling.

When to use:

• Speech recognition to handle different voice pitches
• Music analysis to handle different keys
• Voice cloning and synthesis training

Semitone reference: 12 semitones = 1 octave (doubling/halving frequency)

Constructors

PitchShift(double, double, double, int)

Creates a new pitch shift augmentation.

public PitchShift(double minSemitones = -2, double maxSemitones = 2, double probability = 0.5, int sampleRate = 16000)

Parameters

minSemitones double

Minimum pitch shift in semitones (default: -2.0).

maxSemitones double

Maximum pitch shift in semitones (default: 2.0).

probability double

Probability of applying this augmentation (default: 0.5).

sampleRate int

Sample rate of the audio in Hz (default: 16000).

Properties

MaxSemitones

Gets the maximum pitch shift in semitones.

public double MaxSemitones { get; }

Property Value

double

Remarks

Default: 2.0 semitones (about +12% frequency)

Positive values raise the pitch.

MinSemitones

Gets the minimum pitch shift in semitones.

public double MinSemitones { get; }

Property Value

double

Remarks

Default: -2.0 semitones (about -12% frequency)

Negative values lower the pitch.

Methods

ApplyAugmentation(Tensor<T>, AugmentationContext<T>)

Implement this method to perform the actual augmentation.

protected override Tensor<T> ApplyAugmentation(Tensor<T> data, AugmentationContext<T> context)

Parameters

data Tensor<T>

The input data.

context AugmentationContext<T>

The augmentation context.

Returns

Tensor<T>

The augmented data.

GetParameters()

Gets the parameters of this augmentation.

public override IDictionary<string, object> GetParameters()

Returns

IDictionary<string, object>

A dictionary of parameter names to values.