spectral features added

aclai-lab · Dec 12, 2023 · 137c58c · 137c58c
1 parent 24b15ba
commit 137c58c
Show file tree

Hide file tree

Showing 5 changed files with 279 additions and 34 deletions.
diff --git a/src/featuresExtractor.jl b/src/featuresExtractor.jl
@@ -0,0 +1,80 @@
+include("signalDataStructure.jl")
+include("fft.jl")
+include("mel.jl")
+include("spectral.jl")
+
+function audioFeaturesExtraction(
+    x::AbstractArray{T},
+    sr::Int64;
+    #### define audio objects ####
+    # user defined options
+    frequency_range::Vector{Int64}=[0, Int(round(sr / 2))],
+    numCoeffs::Int=13,
+    melStyle::Symbol=:htk,
+    numBands::Int=32,
+    spectrum_type::Symbol=:power,
+    filterbank_design_domain::Symbol=:linear,
+    filterBankNormalization::Symbol=:bandwidth,
+    window_type::Symbol=:hann,
+    window_length::Int=Int(round(0.03 * sr)),
+    overlap_length::Int=Int(round(0.02 * sr)),
+    rectification::Symbol=:log,
+    logEnergyPos::Symbol=:append,
+    deltaWindowLength::Int=9
+) where {T<:AbstractFloat}
+    # options and data structures definition
+    options = signalSetup(
+        sr=sr,
+        frequency_range=Float64.(frequency_range),
+        numCoeffs=numCoeffs,
+        melStyle=melStyle,
+        numBands=numBands,
+        spectrum_type=spectrum_type,
+        filterbank_design_domain=filterbank_design_domain,
+        filterBankNormalization=filterBankNormalization,
+        window_type=window_type,
+        window_length=window_length,
+        overlap_length=overlap_length,
+        rectification=rectification,
+        logEnergyPos=logEnergyPos,
+        deltaWindowLength=deltaWindowLength
+    )
+
+    # normalize signal
+    x = Float64.(x)
+    x = x ./ maximum(abs.(x))
+
+    data = signalData(
+        x=x
+    )
+
+    takeFFT(data, options)
+    melSpectrogram(data, options)
+    mfcc(data, options)
+    spectral_features(data, options)
+
+    hcat(
+        data.mel_spectrogram',
+        data.coeffs,
+        data.delta,
+        data.deltaDelta,
+        data.spectral_centroid,
+        data.spectral_crest,
+        data.spectral_flatness,
+        data.spectral_flux,
+        data.spectral_decrease,
+        data.spectral_kurtosis,
+        data.spectral_rolloff,
+        data.spectral_skewness,
+        data.spectral_slope,
+        data.spectral_spread
+    )
+end
+
+# debug
+using PyCall
+librosa = pyimport("librosa")
+sr_src = 8000
+x, sr = librosa.load("/home/riccardopasini/Documents/Aclai/Julia_additional_files/test.wav", sr=sr_src, mono=true)
+
+audioFeaturesExtraction(x, sr)
diff --git a/src/fft.jl b/src/fft.jl
@@ -49,7 +49,7 @@ function takeFFT(
     options::signalSetup
 )
     options.FFT_length = options.window_length
-    data.window, unused = gencoswin(options.window_type, options.window_length, :periodic)
+    data.window, unused = gencoswin(options.window_type, options.window_length, :symmetric)
 
     hopLength = options.window_length - options.overlap_length
 
@@ -58,10 +58,10 @@ function takeFFT(
     #     data.window = data.window * options.scale_factor
     # end
 
-    if (options.windowNormalization)
-        options.spectrum_type == :power ? options.scale_factor = sqrt(0.5 * sum(data.window)^2) : options.scale_factor = 0.5 * sum(data.window)
-        data.window = data.window / options.scale_factor
-    end
+    # if (options.windowNormalization)
+    #     options.spectrum_type == :power ? options.scale_factor = sqrt(0.5 * sum(data.window)^2) : options.scale_factor = 0.5 * sum(data.window)
+    #     data.window = data.window / options.scale_factor
+    # end
 
     y = buffer(data.x, options.window_length, hopLength)
 
@@ -70,16 +70,45 @@ function takeFFT(
     E[E.==0] .= floatmin(Float64) # il minimo float al posto di zero
     data.logEnergy = log.(E)
     # apply window
-    data.fft = fft(y .* data.window, (1,))
+    data.fft = (fft(y .* data.window, (1,)))
 
     # Convert to one-sided FFT
-    if (options.oneSided)
-        binHigh = Int(floor(options.FFT_length / 2 + 1))
-        X = data.fft[1:binHigh, :]
-        if (options.spectrum_type == :power)
-            data.fft = real((X .* conj(X)))
-        else
-            data.fft = abs.(X)
-        end
+    # if (options.oneSided)
+    #     binHigh = Int(floor(options.FFT_length / 2 + 1))
+    #     X = data.fft[1:binHigh, :]
+    #     if (options.spectrum_type == :power)
+    #         data.fft = (X .* conj(X))
+    #     else
+    #         data.fft = abs.(X)
+    #     end
+    # end
+
+    # trim to desired range
+    binLow = Int(ceil(options.frequency_range[1] * options.FFT_length / options.sr + 1))
+    binHigh = Int(floor(options.frequency_range[2] * options.FFT_length / options.sr + 1))
+    bins = binLow:binHigh
+    data.fft = data.fft[bins, :]
+    # convert to half-sided magnitude or power spectrum
+    if (options.spectrum_type == :power)
+        data.fft = data.fft .* conj(data.fft) ./ (0.5 * sum(data.window)^2)
+    else # Magnitude
+        data.fft = abs.(data.fft) ./ (0.5 * sum(data.window))
+    end
+    # if the first bin is DC, halve it.
+    if (binLow == 1)
+        data.fft[1, :] = 0.5 * data.fft[1, :]
+    end
+
+    # if the final bin is Nyquist, and FFTLength is even, halve it.
+    if (binHigh == floor(options.FFT_length / 2 + 1) && rem(options.FFT_length, 2) == 0)
+        data.fft[end, :] = 0.5 * data.fft[end, :]
+    end
+
+    # create frequency vector
+    w = ((options.sr / options.FFT_length) .* (collect(bins) .- 1))
+    # shift final bin if fftLength is odd and the final range is full to fs/2.
+    if (rem(options.FFT_length, 2) == 1 && binHigh == floor(options.FFT_length / 2 + 1))
+        w[end] = options.sr * (options.FFT_length - 1) / (2 * options.FFT_length)
     end
+    data.frequency_vector = w[:]
 end # takeFFT(data, options)
diff --git a/src/mel.jl b/src/mel.jl
@@ -57,7 +57,6 @@ end # mel2hz
 
 function designMelFilterBank(
     sr::Int64;
-    oneSided::Bool=true,
     frequencyScale::Symbol=:mel,
     FFT_length::Int64=256,
     numBands::Int64=32,
@@ -70,7 +69,6 @@ function designMelFilterBank(
     # options and data structures definition    
     options = signalSetup(
         sr=sr,
-        oneSided=oneSided,
         frequencyScale=frequencyScale,
         FFT_length=FFT_length,
         numBands=numBands,
@@ -172,10 +170,9 @@ function designMelFilterBank(data::signalData, options::signalSetup)
         end
     end
 
-    if (options.oneSided)
-        select = getOnesidedFFTRange(options.FFT_length)
-        data.filterBank = data.filterBank[:, select]
-    end
+    # take one side
+    select = getOnesidedFFTRange(options.FFT_length)
+    data.filterBank = data.filterBank[:, select]
 end # function designMelFilterBank
 
 function createDCTmatrix(
@@ -247,7 +244,6 @@ function melSpectrogram(
     x::AbstractArray{T},
     sr::Int64;
     frequency_range::Vector{Int64}=[0, Int(round(sr / 2))],
-    oneSided::Bool=true,
     melStyle::Symbol=:htk,
     numBands::Int64=32,
     spectrum_type::Symbol=:power,
@@ -256,7 +252,6 @@ function melSpectrogram(
     window_type::Symbol=:hann,
     window_length::Int64=Int(round(0.03 * sr)),
     overlap_length::Int64=Int(round(0.02 * sr)),
-    windowNormalization::Bool=true,
 ) where {T<:AbstractFloat}
 
     # options and data structures definition    
@@ -268,11 +263,9 @@ function melSpectrogram(
         spectrum_type=spectrum_type,
         filterbank_design_domain=filterbank_design_domain,
         filterBankNormalization=filterBankNormalization,
-        oneSided=oneSided,
         window_type=window_type,
         window_length=window_length,
         overlap_length=overlap_length,
-        windowNormalization=windowNormalization,
     )
 
     data = signalData(
@@ -307,14 +300,12 @@ function mfcc(
     sr::Int64;
     numCoeffs::Int64=13,
     frequency_range::Vector{Int64}=[0, Int(round(sr / 2))],
-    oneSided::Bool=false,
     melStyle::Symbol=:slaney133, # :htk, :slaney, :slaney133
     numBands::Int64=32,
     spectrum_type::Symbol=:power, # :power, :magnitude
     filterbank_design_domain::Symbol=:linear,
     filterBankNormalization::Symbol=:bandwidth, # :bandwidth, :area, :none
     window_type::Symbol=:hann,
-    windowNormalization::Bool=false,
     window_length::Int64=Int(round(0.03 * sr)),
     overlap_length::Int64=Int(round(0.02 * sr)),
     rectification::Symbol=:log, # :log, :cubic
@@ -332,11 +323,9 @@ function mfcc(
         spectrum_type=spectrum_type,
         filterbank_design_domain=filterbank_design_domain,
         filterBankNormalization=filterBankNormalization,
-        oneSided=oneSided,
         window_type=window_type,
         window_length=window_length,
         overlap_length=overlap_length,
-        windowNormalization=windowNormalization,
         rectification=rectification,
         logEnergyPos=logEnergyPos,
         deltaWindowLength=deltaWindowLength
@@ -360,7 +349,7 @@ function mfcc(
 
     # place log energy
     if (options.logEnergyPos == :append)
-        data.coeffs = hcat(data.logEnergy, data.coeffs)
+        data.coeffs = hcat(data.coeffs, data.logEnergy)
     elseif (options.logEnergyPos == :replace)
         # data.coeffs = [logE.',data.coeffs(:,2:end)];
         # da fare
@@ -396,10 +385,8 @@ end
 # coeffs, delta, deltaDelta, loc = mfcc(
 #     x,
 #     sr,
-#     oneSided=true,
 #     melStyle=:htk,
 #     window_length=320,
-#     windowNormalization=true,
 #     overlap_length=240,
 #     logEnergyPos=:none
 # )
diff --git a/src/signalDataStructure.jl b/src/signalDataStructure.jl
@@ -7,7 +7,6 @@ using Parameters
     window_type::Symbol=:hann
     window_length::Int64=0
     overlap_length::Int64=0
-    windowNormalization::Bool=true # possibile cambiarlo in filterBankNormalization?
     normalizationFactor::Float64=1.0
     # mel
     frequency_range::Vector{Float64}=[]
@@ -19,9 +18,8 @@ using Parameters
     frequencyScale::Symbol=:mel
     filterbank_design_domain::Symbol=:linear
     filterBankNormalization::Symbol=:bandwidth # :bandwidth, :area, :none
-    oneSided::Bool=false
     rectification::Symbol=:log
-    logEnergyPos::Symbol=:append
+    logEnergyPos::Symbol=:append #:apend, :replace, :none
     deltaWindowLength::Int64=9
     bandEdges::AbstractVector{AbstractFloat}=[]
 end
@@ -30,6 +28,7 @@ end
     x::AbstractArray{Float64}=[]
     # fft
     fft::AbstractArray{Complex{Float64}}=[]
+    frequency_vector::Vector{Float64}=[]
     window::Vector{Float64}=[]
     # mel
     filterBank::AbstractArray{Float64}=[]
@@ -39,4 +38,15 @@ end
     delta::AbstractArray{Float64}=[]
     deltaDelta::AbstractArray{Float64}=[]
     logEnergy::Vector{Float64}=[]
+    # spectral
+    spectral_centroid::Vector{Float64}=[]
+    spectral_crest::Vector{Float64}=[]
+    spectral_flatness::Vector{Float64}=[]
+    spectral_flux::Vector{Float64}=[]
+    spectral_decrease::Vector{Float64}=[]
+    spectral_kurtosis::Vector{Float64}=[]
+    spectral_rolloff::Vector{Float64}=[]
+    spectral_skewness::Vector{Float64}=[]
+    spectral_slope::Vector{Float64}=[]
+    spectral_spread::Vector{Float64}=[]
 end