fastmath.transform — generateme/fastmath 3.0.0-alpha2

Transforms.

See transformer and [[TransformProto]] for details.

Wavelet

Based on JWave library.

Be aware that some of the wavelet types doesn't work properly. :battle-23, :cdf-53, :cdf-97.

Cos/Sin/Hadamard

Orthogonal or standard fast sine/cosine/hadamard 1d transforms.

Fourier

DFT, FFT, DHT.

compress clj

(compress xs mag)

(compress trans xs mag)

Compress transformed signal xs with given magnitude mag.

compress-peaks-average clj

(compress-peaks-average xs)

(compress-peaks-average trans xs)

Compress transformed signal xs with peaks average as a magnitude

denoise clj

(denoise xs)

(denoise xs
         {:keys [method threshold skip]
          :or {method :hard threshold :universal skip 0}})

(denoise trans xs method)

Wavelet shrinkage with some threshold.

Methods can be:

• :hard (default)
• :soft
• :garrote
• :hyperbole

:threshold can be a number of one of the denoise-threshold methods (default: :visu)

:skip can be used to leave :skip number of coefficients unaffected (default: 0)

Use on transformed sequences or call with transformer object.

denoise-threshold clj

(denoise-threshold xs threshold)

Calculate optimal denoise threshold.

threshold is one of the following

• :visu - based on median absolute deviation estimate (default)
• :universal - based on standard deviation estimate
• :sure or :rigrsure - based on SURE estimator
• :hybrid or :heursure - hybrid SURE estimator

forward-1d clj

(forward-1d t xs)

Forward transform of sequence or array.

forward-2d clj

(forward-2d t xss)

Forward transform of sequence or array.

reverse-1d clj

(reverse-1d t xs)

Forward transform of sequence or array.

reverse-2d clj

(reverse-2d t xss)

Forward transform of sequence or array.

transformer clj multimethod

Create transform object for given wavelet.

Wavelets

• :fast for 1d or 2d Fast Wavelet Transform. Size of data should be power of 2.
• :packet for 1d or 2d Wavelet Packet Transform. Size of data should be power of 2.
• :decomposed-fast for 1d Fast Wavelet Transform. Data can have any size (Ancient Egyptian Decomposition is used).
• :decomposed-packet for 1d Wavelet Packet Transform. Data can have any size (Ancient Egyptian Decomposition is used).

Second argument is wavelet name as key. See wavelets-list for all supported names.

Sine/Cosine/Hadamard

• :standard for 1d :sine, :cosine, :hadamard.
• :orthogonal for 1d :sine, :cosine.

Note that :sine and :cosine require first element to be equal 0. Size of data should be power of 2.

Fourier

• :standard :dft - 1d Discrete Fourier Transform - returns double-array where even elements are real part, odd elements are imaginary part.

wavelets-list clj

List of all possible wavelets.