SincMaths

SincMaths is a Kotlin Multiplatform library which provides a 2D matrix SincMatrix to facilitate translation of signal processing code written in Octave/MATLAB to mobile applications. Originally written for the Gait&Balance app. See docs for a comprehensive list of functions. Key implementation aspects:

• Stores data in row-major format as opposed to column-major format used by Octave/MATLAB
• Uses ejml-simple for Android side optimisations
• Uses Apple Accelerate for iOS side optimisations
• Uses rafat/wavelib to provide wavelet-based demonising
• Uses codeplea/tinyexpr to provide Octave/MATLAB style indexing

Examples

Initialisation

val matrixA = matrixFrom("[1, 2, 3; 4, 5, 6]")
val matrixB = matrixFrom("[1, 2, 3, 4]")
val matrixC = matrixFrom("[5;6;7;8;9;10]")
val matrixD = matrixFrom("1:10")
val matrixE = matrixFrom("-1.5:-1:-7.9")

val matrixF = matrixOf(2, 10, 1..20)
val matrixG = colVectorOf(1.0, 0.5, 2.9, 10.1, 15.4)

Indexing

val matrixM: SincMatrix = (1..110).asSincMatrix(11, 10)

val matrixA = matrixM["1:5,4:7"]
val matrixB = matrixM["1:end,end:-1:end-1"]

val matrixC = matrixM[3, 3]
val matrixD = matrixM[1]

val matrixE = matrixM[1..4]

val matrixF = matrixM.get { endR, endC, allR, allC ->
    Pair(allR, 4..7)
} // same as matrixM[":,4:7"]

Implementation of acf from MATLAB Central

Reference: https://au.mathworks.com/matlabcentral/fileexchange/30540-autocorrelation-function-acf

fun SincMatrix.acf(numLags: Int): SincMatrix {

    require(this.isVector) { "This function works only for vectors" }
    require(numLags < this.numel) {
        "No. of lags should be smaller than the length of the vector"
    }

    val zeroMeanVector = this - this.mean().scalar
    val convSum = zeroMeanVector.conv(bVector = zeroMeanVector.flip())
    val scale = 1.0 / zeroMeanVector.dot(zeroMeanVector).scalar
    val scaledConvSum = convSum * scale
    val acfElements = this.numel + 1..this.numel + numLags
    return if (this.isRow) {
        scaledConvSum[this.rowIndicesRange, acfElements]
    } else {
        scaledConvSum[acfElements, this.colIndicesRange]
    }
}

Usage

Kotlin Project: MavenCentral

Note: This library does not work with iosSimulatorArm64 target. Android tests have to be performed on an emulator or a connected device, otherwise, the underlying JAVA JNI libraries fail to load.

Add dependency:

implementation("io.github.gallvp:sincmathsmp:0.3")

Try a test:

expect(5050.0) {
    matrixFrom("1:100").sum().scalar
}

Swift Project: CocoaPods

Note: Global functions such as matrixFrom are converted into open class functions as SincMatrixInitKt.matrixFrom. Use 'Jump to Definition' on 'SincMaths' to locate the class and function definitions.

Clone this project and add it to your project's Podfile:

pod 'SincMaths', :path => '/path/to/cloned/sincmaths/sincmathsmp'

Import it in your project and try a test:

import SincMaths

XCTAssertEqual(
    SincMatrixInitKt.matrixFrom(script: "1:100").sum(dim: 1).scalar,
    5050.0
)