paradigma 0.3.2__py3-none-any.whl → 0.4.0__py3-none-any.whl

paradigma/tremor/TremorFeaturesAndClassification.m

@@ -1,345 +0,0 @@
-% #########################################################################
-% Input
-% Conf - Structure
-%     .WindowSizeTime - scalar - double - time of the window in seconds
-%     .Fs             - scalar - double - sampling frequency in Hz
-%
-% Data_In - cell      - array - double - Nsamples x Nchannels - IMU signal in
-%                                  columns [x y z]
-%
-% t - cell            - array - double - Nsamples x 1 - IMU time stamp - 
-%                                  Unix time corrected (interpolated) 
-%
-% Type of the signal  - string - IMU type of signal 'Gy' - gyroscope
-%                                                   'Ac' - acceleration
-%
-% #########################################################################
-% Output
-% Features - structure
-%         .WindowIniTime  - array - double - Nwindows x 1 - Unix Corrected Time (after interpolation) -
-%                                                           Initial time stamp of the window
-%
-%         .Derivatives   - table - double - Nwindows x 3 - Mean angular velocity absolute derivatives for each axis (columns):
-%                                             .GyMeanDx - x-axis
-%                                             .GyMeanDy - y-axis
-%                                             .GyMeanDz - z-axis
-%
-%         .PowerAxis       - table - double - Nwindows x 3 - Full band power axis (0.5-25 Hz) (columns):
-%                                             .GyPowerX - x-axis
-%                                             .GyPowerY - y-axis   
-%                                             .GyPowerZ - z-axis
-%
-%         .DomTremorPowerAxis       - table - double - Nwindows x 3 - Tremor dominant power (3-7 Hz) for each axis (columns):
-%                                             .GyLTreDomPowerX - x-axis
-%                                             .GyLTreDomPowerY - y-axis   
-%                                             .GyLTreDomPowerZ - z-axis
-%
-%         .FreqPeak      - table - double - Nwindows x 1 - Dominant
-%         frequency in the PSD summed spectrum (PSD = PSDx + PSDy + PSDz) within the range 1-25Hz per
-%         window
-%                                             .GyFreqPeak - frequency of
-%                                             the PSD peak
-
-%         .PSDTremorFeatures  - table - double - Nwindows x 3 - PSD-based
-%         featured extracted in the tremor range (3-7) Hz per window 
-%         - orientation invariant (PSD = PSDx + PSDy + PSDz)
-%                                             .GyLTreBandPower  - power in the tremor range
-%                                             .GyLTreFreqPeak   - dominant frequency in the tremor range
-%                                             .GyLTreDomPower   - dominant power 1.25 Hz around the tremor peak   
-%
-%         .PSDHighTremorFeatures  - table - double - Nwindows x 3 - PSD-based
-%         featured extracted in the high tremor range (7-12) Hz per window
-%         - orientation invariant (PSD = PSDx + PSDy + PSDz)
-%                                             .GyHTreBandPower - power in the high tremor range
-%                                             .GyHTreFreqPeak  - dominant frequency in the high tremor range
-%                                             .GyHTreDomPower  - dominant power 1.25 Hz around the high tremor peak   
-%
-%         .MelCepsCoeff  - table - double - Nwindows x 12 -
-%         Mel-Frequency cepstral coefficients from 1 to 12 per window -
-%         orientation invariant (sum spectrograms x, y, z)
-%                                             .GyMFCC1   - 1st MFCC coefficient
-%                                             .GyMFCC2   - 2nd  MFCC coefficient
-%                                             ...
-%                                             .GyMFCC12  - 12th MFCC coefficient
-%
-%
-%         .PowerArmActv - table - double Nwindows x 1 - Power in the Arm Activity
-%                                          frequency range (0.5 - 3 Hz)
-%                                          .GyArmActvPower - power in the range (0.5 - 3 Hz)
-%
-%
-%         .PowerHighFreq - table - double Nwindows x 1 - Power in the high frequency domain (12-25 Hz)
-%                                          .GyHighFreqPower - power in the
-%                                          range (12-25 Hz)
-% 
-%         .TremorProb - array - double - Nwindows x 1 - tremor
-%         probability obtained by logistic regression based on MFCCs
-%         features
-% 
-%         .TremorHat  - array - double - Nwindows x 1    - tremor label obtained after applying threshold in the tremor probability 
-%         .RestTremorHat - array - double - Nwindows x 1 - rest tremor
-%         label after logical AND operation between TremorHat and SpectralCheck (PSD peak falls within tremor domain [3 - 7 Hz])
-% #########################################################################
-
-function [Features] = TremorFeaturesAndClassification(Conf,Data_In,t,TypeOfSignal)
-
-% Load Classifiers obtained from PD@Home
-load('TremorClassifier2.mat'); % Load the updated tremor model
-load('Threshold_total.mat');   % Load the threshold
-
-TremorThr = Threshold_total;   % Threshold at 95% specificity
-
-load('MeanVectorPDhome.mat');  % Scaling parameters
-load('SigmaVectorPDhome.mat'); % Scaling parameters
-
-% Getting parameters
-Fs = Conf.Fs;                           % Sampling frequency
-WindowSizeTime = Conf.WindowSizeTime;   % window interval in seconds
-
-% Evaluating slicing parameters
-DeltaSample = floor(WindowSizeTime*Fs);               % number of samples for skipping to get the next window
-                                                      % Introduce overlap between samples if needed
-NumWindows = floor(length(Data_In(:,1))/DeltaSample); % number of windows given the segment size (Data_In)
-
-% Initialize variables
-WindowIniTime      = [];
-DomTremorPowerAxis = [];
-PowerAxis          = [];
-
-MelCepsCoeff  = [];
-Derivatives   = [];
-PSDTremorFeatures     = [];
-PSDHighTremorFeatures = []; 
-FreqPeak      = [];
-ArmActvPower  = [];
-HighFreqPower = [];
-
-TremorProb    = [];
-TremorHat     = [];
-PeakFreqTremorCheck = [];
-RestTremorHat = [];
-
-% Loop across the windows
-for kk = 1:NumWindows  
-
-    % Get the inital time stamp
-    WindowIniTime(kk)        = t((kk-1)*DeltaSample+1);
-
-    % Slice data according WindowSizeTime
-    sig_x  = Data_In(((kk-1)*DeltaSample+1):kk*DeltaSample,1);
-    sig_y  = Data_In(((kk-1)*DeltaSample+1):kk*DeltaSample,2);
-    sig_z  = Data_In(((kk-1)*DeltaSample+1):kk*DeltaSample,3);
-
-    DataMatrix = [sig_x sig_y sig_z];                               % Concatenating Data - Signal Matrix           
-
-    %% PSD estimation - Welch periodogram function parameters - Evaluate the Spectrum
-    % Configuration for power spectrum density (PSD) estimation
-    ConfPSD.Fs        = Fs;        % sampling frequency in Hz
-    ConfPSD.pwelchwin = 3*Fs;      % number of samples for the subwindows used in the periodogram
-    ConfPSD.noverlap  = round(0.8*ConfPSD.pwelchwin);  % number of samples for overlapping subwindows 
-    ConfPSD.nfft      = 4*Fs;      % NFFT number of samples - keep spectral resolution in 0.25 Hz
-    ConfPSD.freqmax   = 25;        % Maximum frequency to be considered
-    ConfPSD.freqmin   = 1;         % Lower frequency limit used in PSDest
-                                   % for estimating the dominant frequency in
-                                   % the range [freqmin freqmax] Hz. This low
-                                   % frequency cut is due to gyroscope drift and the presence of enhanced low frequency components in this case.
-    
-    [PSDMatrix,FreqVect,FreqPeak(kk)] = PSDEst(ConfPSD,DataMatrix); % Get the PSD Matrix
-    
-    % Check extra criterion for rest tremor: Does the dominant peak within
-    %                                        1-25 Hz falls into the tremor range (3-7) Hz
-    
-    PeakFreqTremorCheck(kk) = (FreqPeak(kk) >= 3 & FreqPeak(kk) <= 7);
-
-    %% Mel-Cepstrum Coefficients Feature Extraction
-    % Configuration for MFCCs extraction
-    ConfMFCC.Fs            = Fs;    % Sampling Frequency
-    ConfMFCC.numFilters    = 15;    % Number of filters to be used in the Mel-scale
-    ConfMFCC.NumMelCoeff   = 12;    % number of MFCCs coefficients
-    ConfMFCC.StrAxis       = 'Sum'; % Evaluation based on a summed spectrogram, i.e., orientation independent
-    ConfMFCC.MaxFreqFilter = 25;    % Maximal frequency in the bank filter
-    ConfMFCC.MFCCwin       = 2;     % Subwindow in seconds for evaluating the spectrogram
-
-    [MelCepsCoeffAux] = MFCCExtract(ConfMFCC,DataMatrix); % Perform MFCC extraction
-
-    %% Tremor classification
-    DataScaled = (MelCepsCoeffAux - MeanVector)./SigmaVector; % Scale MFCCs features for classification
-    TremorProb(kk)  = predict(Mdl,DataScaled);                % Tremor probability based on logistic regression classifier
-    
-    TremorHat(kk)      = (TremorProb(kk) > TremorThr); % Boolean
-    RestTremorHat(kk)  = (TremorProb(kk) > TremorThr) & (PeakFreqTremorCheck(kk)); % Boolean
-
-    MelCepsCoeff = vertcat(MelCepsCoeff,MelCepsCoeffAux);    % Concatenate values
-    clear('MelCepsCoeffAux'); % clear temp variable
-
-    %% Tremor Severity and arm activity features (step 5 for eScience toolbox)
-
-    % Derivatives Evaluation - Get the mean absolute derivative values
-    DerivativesAux   = DerivativesExtract(sig_x,sig_y,sig_z,Fs); % Evaluate mean signals' absolute derivatives values
-    Derivatives      = vertcat(Derivatives,DerivativesAux);         % Concatenate values - store Derivatives
-    clear('DerivativesAux');                                        % Clear the auxiliar strucuture
-
-    % Parameters for obtaning spectral features
-    fSpectResol       = Fs/ConfPSD.nfft;                       % Spectral resolution: fixed Fs/(4*Fs) = 0.25 Hz
-    ConfPSDExtr.SpectralIntervalPeak = round(0.5/fSpectResol); % The interval for calculating the dominant power
-    % is 0.5 Hz for each side of the peak,
-    % which implies in different number of
-    % samples in the freq domain depending
-    % on the spectral resolution
-
-    ConfPSDExtr.StrAxis = 'Sum';         % PSD features will be evaluated in the PSDsum domain, i.e., PSDx+PSDy+PSDz: orientation independent
-    ConfPSDExtr.nfft    = 4*Fs;          % Number of points FFT
-    ConfPSDExtr.Fs      = Fs;            % Sampling frequency
-    ConfPSDExtr.MaxFreqOverallBand = 25; % Maximal frequency for evaluating total power
-
-    % Low Tremor Features
-    ConfPSDExtr.freqrange = [3 7];       % IMPORTANT - define here the bandwidth for tremor band
-    [PSDTremorFeaturesAux] = PSDExtrOpt(ConfPSDExtr,PSDMatrix,FreqVect);     % Get Low Freq Tremor Features
-    PSDTremorFeatures      = vertcat(PSDTremorFeatures,PSDTremorFeaturesAux);   % Store low freq tremor features
-    clear('PSDTremorFeaturesAux');                                              % Clear temp variable
-
-    % In the tremor range: get PSD Dominant power for each axis
-    
-    ConfAxis.StrAxis = 'X';
-    ConfAxis.freqrange = [3 7];       % Frequency range to get tremor dominant power
-    ConfAxis.nfft    = 4*Fs;          % Number of points FFT
-    ConfAxis.Fs      = Fs;            % Sampling frequency
-    ConfAxis.SpectralIntervalPeak = round(0.5/fSpectResol); % The interval for calculating the dominant power
-    ConfAxis.MinFreqOverallBand = 0.5; % Minimal frequency for evaluating total power
-    ConfAxis.MaxFreqOverallBand = 25;  % Maximal frequency for evaluating total power
-    
-    [PSD_X]  = PSDExtrAxis(ConfAxis,PSDMatrix,FreqVect);
-    
-    ConfAxis.StrAxis = 'Y';
-    [PSD_Y]  = PSDExtrAxis(ConfAxis,PSDMatrix,FreqVect);
-    
-    ConfAxis.StrAxis = 'Z';
-    [PSD_Z]  = PSDExtrAxis(ConfAxis,PSDMatrix,FreqVect);
-
-    % Get the fixed bandwith dominant power 1.25 Hz around the peak in the
-    % tremor range
-    
-    PSDPowerAxisAux               = [PSD_X(1) PSD_Y(1) PSD_Z(1)];
-    PSDTremorDomPowerAxisAux      = [PSD_X(2) PSD_Y(2) PSD_Z(2)];
-    
-    PowerAxis                  = vertcat(PowerAxis,PSDPowerAxisAux);
-    DomTremorPowerAxis         = vertcat(DomTremorPowerAxis,PSDTremorDomPowerAxisAux);
-    
-    clear('PSD_X','PSD_Y','PSD_Z','PSDTremorDomPowerAxisAux','PSDPowerAxisAux');
-
-    % Get features in the high tremor range
-    ConfPSDExtr.freqrange = [7 12];                                              % IMPORTANT - define here the bandwidth 
-    [PSDHighTremorFeaturesAux] = PSDExtrOpt(ConfPSDExtr,PSDMatrix,FreqVect);  % Get High Freq Tremor Features
-    PSDHighTremorFeatures      = vertcat(PSDHighTremorFeatures,PSDHighTremorFeaturesAux);   % Store
-    clear('PSDHighTremorFeaturesAux');                                                      % Clear temp variable
-
-    % Get Power in the arm-activity
-    ConfBP.freqrange = [0.5 3];    % Frequency range   
-    ConfBP.StrAxis = 'Sum'; % Axis or sum
-    ConfBP.nfft = 4*Fs;     % number of samples for FFT
-    ConfBP.Fs = Fs;         % Sampling frequency
-    
-    [ArmActvPowerAux] = PSDBandPower(ConfBP,PSDMatrix,FreqVect);     % Get arm-activity bandpower
-    ArmActvPower      = vertcat(ArmActvPower,ArmActvPowerAux);       % Store 
-    clear('ArmActvPowerAux');                                        % Clear temp variable
-
-    % Get Power in high frequency band 
-    ConfBP.freqrange = [12 25];       % Frequency range
-    [HighFreqPowerAux] = PSDBandPower(ConfBP,PSDMatrix,FreqVect); % Get high frequency bandpower
-    HighFreqPower      = vertcat(HighFreqPower,HighFreqPowerAux); % Store 
-    clear('HighFreqPowerAux');                                    % Clear temp variable
-   
-end
-% Organizing Output as a table
-
-%% Frequency Peak
-VarName = {'GyFreqPeak'};
-FreqPeak = array2table(FreqPeak','VariableNames',VarName);
-
-%% MFCCs
-% Creating MFCCs names to be used
-for pp = 1:ConfMFCC.NumMelCoeff
-    CepsName{1,pp} = [TypeOfSignal,'MFCC',num2str(pp)];
-end
-
-VarName = CepsName;
-% Defining the output
-MelCepsCoeff = array2table(MelCepsCoeff,'VariableNames',VarName);
-clear('VarName');
-
-%% Derivatives
-VarName = [{[TypeOfSignal,'MeanDx']},{[TypeOfSignal,'MeanDy']},{[TypeOfSignal,'MeanDz']}];
-
-% Converting array to table
-Derivatives = array2table(Derivatives,"VariableNames",VarName);
-clear('VarName');
-
-%% PSD Features Arm Activity Band
-VarName      = {[TypeOfSignal,'ArmActvPower']};
-ArmActvPower = array2table(ArmActvPower,"VariableNames",VarName);
-clear('VarName');
-
-%% PSD Features High Frequency Band
-VarName = {[TypeOfSignal,'HighFreqPower']};
-HighFreqPower = array2table(HighFreqPower,"VariableNames",VarName);
-clear('VarName');
-
-% Tremor Band
-StrBand   = 'LTre';                 % String specifying the bandwidth in 
-                                    % the obtained features name
-
-VarName = {[TypeOfSignal,StrBand,'BandPower'], ...
-           [TypeOfSignal,StrBand,'FreqPeak'],...
-           [TypeOfSignal,StrBand,'DomPower']};
-
-PSDTremorFeatures = array2table(PSDTremorFeatures,'VariableNames',VarName);
-clear('VarName');
-
-%% Dominant tremor power in each axis
-VarName = {[TypeOfSignal,StrBand,'DomPowerX'],...
-           [TypeOfSignal,StrBand,'DomPowerY'],...
-           [TypeOfSignal,StrBand,'DomPowerZ']};
-
-DomTremorPowerAxis  = array2table(DomTremorPowerAxis ,'VariableNames',VarName);
-clear('VarName');
-
-%% Full band: (0.5-25 Hz) in each axis
-VarName = {[TypeOfSignal,'PowerX'],...
-           [TypeOfSignal,'PowerY'],...
-           [TypeOfSignal,'PowerZ']};
-
-PowerAxis  = array2table(PowerAxis ,'VariableNames',VarName);
-clear('VarName');
-
-%% Features in High Tremor Band (7 - 12 Hz)
-StrBand   = 'HTre';      % String specifying the bandwidth in the obtained features name
-VarName = {[TypeOfSignal,StrBand,'BandPower'], ...
-           [TypeOfSignal,StrBand,'FreqPeak'],...
-           [TypeOfSignal,StrBand,'DomPower']};
-
-PSDHighTremorFeatures = array2table(PSDHighTremorFeatures,'VariableNames',VarName);
-clear('VarName');
-
-if NumWindows == 0 % If there are not enough samples for having at least one window, the output structure is empty
-    
-    Features = [];
-else % otherwise, Features receive the concatenated arrays and tables evaluated for the windows
-
-    Features.WindowIniTime = WindowIniTime';
-    Features.Derivatives   = Derivatives;
-    Features.PowerAxis     = PowerAxis; 
-    Features.DomTremorPowerAxis  = DomTremorPowerAxis;
-    
-    Features.FreqPeak      = FreqPeak;
-    Features.PSDTremorFeatures  = PSDTremorFeatures;
-    Features.PSDHighTremorFeatures  = PSDHighTremorFeatures;
-    
-    Features.MelCepsCoeff  = MelCepsCoeff;
-    Features.PowerArmActv       = ArmActvPower;
-    Features.PowerHighFreq      = HighFreqPower;
-    
-    Features.TremorProb    = TremorProb';
-    Features.TremorHat     = TremorHat';
-    Features.RestTremorHat = RestTremorHat';
-
-end

paradigma/tremor/feat_extraction/DerivativesExtract.m

@@ -1,22 +0,0 @@
-% Feature Extraction
-% Estimate the mean absolute derivative values for a given IMU axes window
-% #########################################################################
-% Input:
-% x - x-axis signal (array) - double
-% y - y-axis signal (array) - double
-% z - z-axis signal (array) - double
-% Fs - Sampling frequency (scalar) - double
-%
-% #########################################################################
-% Output:
-% DerivativesOut - array - double
-% [MeanAbsDx MeanAbsDy MeanAbsDz]: 1 x 3
-% #########################################################################
-function DerivativesOut = DerivativesExtract(x,y,z,Fs)
-
-Ts = 1/Fs; % sampling step
-dx = mean((1/Ts)*abs(diff(x,1,1))); % get the mean dx/dt
-dy = mean((1/Ts)*abs(diff(y,1,1))); % get the mean dy/dt
-dz = mean((1/Ts)*abs(diff(z,1,1))); % get the mean dz/dt
-
-DerivativesOut = [dx dy dz]; % Defining output values

paradigma/tremor/feat_extraction/ExtractBandSignalsRMS.m

@@ -1,72 +0,0 @@
-function [RMSValuesOut] = ExtractBandSignalsRMS(Conf,Data,TremorHatLabel)
-
-% Input ###################################################################
-% Conf - Configuration Structure
-%     .Fs - scalar - double - sampling frequency
-%
-% Data         - matrix (Nsamples x Nsignals) - double - columns: IMU signals ([x y and z])
-%
-% TremorHatLabel - scalar - double - rest tremor label
-%
-% Output
-% RMSValuesOut - matrix - double - (1 x 7) .'TypeOfSignal'TremorRMSx   - root mean square value in the tremor range (3 - 7) Hz - x axis
-%                                          .'TypeOfSignal'TremorRMSy   - root mean square value in the tremor range (3 - 7) Hz - y axis
-%                                          .'TypeOfSignal'TremorRMSz   - root mean square value in the tremor range (3 - 7) Hz - z axis
-%                                          .'TypeOfSignal'SumTremorRMS - root mean square value in the tremor range (3 - 7) Hz - sum: (RMSx + RMSy + RMSz)
-%                                          .'TypeOfSignal'SumGaitRMS   - root mean square value in the gait range (0.4 - 2) Hz - sum: (RMSx + RMSy + RMSz)
-%                                          .'TypeOfSignal'SumHighTremorRMS - root mean square value in the high tremor range (7 - 12) Hz - sum: (RMSx + RMSy + RMSz)
-%                                          .'TypeOfSignal'VeryHighBandRMS  - root mean square value in the high tremor range (12- 20) Hz  - sum: (RMSx + RMSy + RMSz)
-% #########################################################################
-
-if TremorHatLabel == 1
-
-    Fs = Conf.Fs; % Sampling frequency
-
-    % Load filter coefficients previously designed and saved in matlab
-    % 6th order bandpass Butterworth filters
-
-    switch Fs % Check the sampling frequency and load the corresponding filter
-        case 100
-            load('BP_Butter_filter_0_4_2_fs_100Hz.mat');
-            load('BP_Butter_filter_3_7_fs_100Hz.mat');
-            load('BP_Butter_filter_7_12_fs_100Hz.mat');
-            load('BP_Butter_filter_12_20_fs_100Hz.mat');
-
-        case 50
-
-            load('BP_Butter_filter_0_4_2_fs_50Hz.mat');
-            load('BP_Butter_filter_3_7_fs_50Hz.mat');
-            load('BP_Butter_filter_7_12_fs_50Hz.mat');
-            load('BP_Butter_filter_12_20_fs_50Hz.mat');
-    end
-    %% Filtering on the required ranges
-    FilGaitBand       = filtfilt(SOSGait,GGait,Data);
-    FilTremorBand     = filtfilt(SOSTre,GTre,Data);
-    FilHighTremorBand = filtfilt(SOSHTre,GHTre,Data);
-    FilVeryHighBand   = filtfilt(SOSVHTre,GVHTre,Data);
-
-    %% Evaluating RMS value
-    GaitRMS         = rms(FilGaitBand); % returns a vector [RMSx RMSy RMSz]
-    TremorRMS       = rms(FilTremorBand);
-    HighTremorRMS   = rms(FilHighTremorBand);
-    VeryHighBandRMS = rms(FilVeryHighBand);
-
-    % Sum across axes
-    SumGaitRMS         = sum(GaitRMS);
-    SumTremorRMS       = sum(TremorRMS);
-    SumHighTremorRMS   = sum(HighTremorRMS);
-    SumVeryHighBandRMS = sum(VeryHighBandRMS);
-
-    % Getting the RMS in the tremor range for each axis to be included in the output
-    TremorRMSx = TremorRMS(1);
-    TremorRMSy = TremorRMS(2);
-    TremorRMSz = TremorRMS(3);
-
-    % Concatenate the features
-    RMSValuesOut = [TremorRMSx  TremorRMSy TremorRMSz SumTremorRMS SumGaitRMS SumHighTremorRMS SumVeryHighBandRMS];
-
-else
-
-    RMSValuesOut = NaN(1,7);
-
-end

paradigma/tremor/feat_extraction/MFCCExtract.m

@@ -1,100 +0,0 @@
-function [MFCCFeatureOut] = MFCCExtract(Conf,Data)
-% Input ###################################################################
-% Conf - Structure
-%     .Fs            - scalar - double - sampling frequency in Hz
-%     .numFilters    - scalar - double -  number of filters in the mel scale
-%     .NumMelCoeff   - scalar - double -  number of MFCCs coefficients
-%     .StrAxis       - string - specifying axis (or sum) to be considered for evaluating the spectrogram
-%     .MaxFreqFilter - scalar - double - maximal frequency (Hz) to be considred in the bank filter
-%     .MFCCwin       - scalar - double - subwindow in seconds adopted for evaluating the spectrogram
-%
-% Data - matrix (Nsamples x Nsignals) - double - columns: IMU signals ([x y and z])
-%
-%
-% Output
-% MFCCFeatureOut - matrix - double - .'TypeOfSignal'MFCC'coefficient number'
-% #########################################################################
-
-[Nsamples,Nsignals] = size(Data); % get number of samples and number of signals
-
-Fs          = Conf.Fs;          % sampling rate
-numFilters  = Conf.numFilters;  % number of filters
-NumMelCoeff = Conf.NumMelCoeff; % number of MFCCs coefficients
-StrAxis     = Conf.StrAxis;     % axis or sum of axes to be considered in the spectrogram
-
-filterbankStart = 0;                  % frequency for starting the bank filter
-filterbankEnd   = Conf.MaxFreqFilter; % maximum frequency in the bank filter
-
-numBandEdges = numFilters + 2;         % number of filter edges
-
-NFFT         = Conf.MFCCwin*Fs;        % number of samples for FFT = number of samples of the adopted subwindow to evaluate the spectrogram
-
-fresol       = Fs/NFFT;                % frequency resolution if 0.5 Hz
-
-filterBank   = zeros(numFilters,NFFT); % initializing the bank filter
-                
-x = filterbankStart:0.01:filterbankEnd; % linear spacing in the bank filter frequency range
-
-melscale = 64.875*log10( 1 + x./17.5); % creating mel scale
-
-xlinmel = linspace(melscale(1),melscale(end),numBandEdges); % equally intervals in the mel scale
-
-bandEdges = 17.5*(10.^(xlinmel/64.875) - 1);    % getting the edges in the frequency domain
-
-bandEdgesBins = round((bandEdges/Fs)*NFFT) + 1; % rouding the edges
-
-% Getting each filter in the frequency domain
-for ii = 1:numFilters
-    filt = triang(bandEdgesBins(ii+2)-bandEdgesBins(ii)); % Triangular filters based on the edges
-    leftPad = bandEdgesBins(ii);             % introduce zero padding on the left
-    rightPad = NFFT - numel(filt) - leftPad; % introduce zero padding on the right
-    filterBank(ii,:) = [zeros(1,leftPad),filt',zeros(1,rightPad)]; % concatenate the vector
-    filterBank(ii,:) = filterBank(ii,:)/(fresol*sum(filterBank(ii,:))); % Normalize the filter coefficients by the bandpass area
-end
-
-% Check the filter profiles if needed
-% frequencyVector = (Fs/NFFT)*(0:NFFT-1);
-% plot(frequencyVector,filterBank');
-% xlabel('Hz')
-% axis([0 frequencyVector(NFFT/2) 0 1])
-
-overlap = round(0.8*NFFT); % ovelap between each subwindow for evaluating the spectrogram
-
-% Extracting the MFCC according to axis (or sum) specified when evaluating
-% the spectrogram
-switch StrAxis
-    case 'X'
-        Signal = Data(:,1); % Get the signal
-        % Evaluate the spectrogram
-        [S1,f,t] = stft(Signal,Fs,"Window",hann(NFFT,'periodic'),...
-            'OverlapLength',overlap,"FrequencyRange","twosided");
-        S = abs(S1); % Get the absolute value
-    case 'Y'
-        Signal = Data(:,2);
-        [S2,f,t] = stft(Signal,Fs,"Window",hann(NFFT,'periodic'),...
-            'OverlapLength',overlap,"FrequencyRange","twosided");
-        S = abs(S2);
-    case 'Z'
-        Signal = Data(:,3);
-        [S3,f,t] = stft(Signal,Fs,"Window",hann(NFFT,'periodic'),...
-            'OverlapLength',overlap,"FrequencyRange","twosided");
-        S = abs(S3);
-    case 'Sum'
-        [S1,f,t] = stft(Data(:,1),Fs,"Window",hann(NFFT,'periodic'),...
-            'OverlapLength',overlap,"FrequencyRange","twosided");
-        [S2,f,t] = stft(Data(:,2),Fs,"Window",hann(NFFT,'periodic'),...
-            'OverlapLength',overlap,"FrequencyRange","twosided");
-        [S3,f,t] = stft(Data(:,3),Fs,"Window",hann(NFFT,'periodic'),...
-            'OverlapLength',overlap,"FrequencyRange","twosided");
-        S = abs(S1) + abs(S2) + abs(S3); % Get the sum of the absolute values across axes 
-end
-
-% Applying the filter bank
-Spec      = filterBank*S;
-
-% Obtaning the cepstral coefficients
-CepsCoeff = cepstralCoefficients(Spec,'NumCoeffs',NumMelCoeff);
-
-Ceps = mean(CepsCoeff); % Taking the average of the cepstral coefficients across time
-
-MFCCFeatureOut = Ceps; % Defining the output

paradigma/tremor/feat_extraction/PSDBandPower.m

@@ -1,52 +0,0 @@
-function  [PSDBandPowerOut] = PSDBandPower(Conf,Data,freq)
-% #########################################################################
-% Input
-% Conf - Structure
-%     .freqrange            - array - double - Frequency Range for feature Extraction
-%     .StrAxis              - string - signal axis (or Sum) specified to be considered for feature extraction
-%     .nfft                 - scalar - double - number of points for the
-%                             FFT (number of samples)
-%     .Fs                   - scalar - double - sampling Rate
-%
-% Data - PSD Matrix - matrix - double - N frequency values x (Nsignals + 1) - last column is the PSD sum (PSDx + PSDy + PSDz)
-%
-% freq - array - double - frequency vector associsted to the PSD;
-%
-%
-%
-% Output - Table with PSD features for the axis (or sum of the axis)
-% selected
-% PSDFeaturesOut - scalar - double (1 x 1)
-%                    BandPower      - power in the range
-
-% Get the estimation parameters
-freqrange =  Conf.freqrange;
-StrAxis   = Conf.StrAxis;
-nfft      = Conf.nfft;
-Fs        = Conf.Fs;
-
-% Getting the signal axis or sum according to what specified in the input
-switch StrAxis
-    case 'X'
-        Pxx = Data(:,1);
-    case 'Y'
-        Pxx = Data(:,2);
-    case 'Z'
-        Pxx = Data(:,3);
-    case 'Sum'
-        Pxx = Data(:,4);
-end
-
-f0 = Fs/nfft; % Spectral Resolution
-
-FreqLowerBound   = freqrange(1); % lower frequency bound
-FreqUpperBound   = freqrange(2); % upper frequency bound
-
-% Total Power - area under PSD using an approximation by the left
-
-InitialIndexBandForPower = find(freq == FreqLowerBound);      % Initial index for bandpower estimation
-FinalIndexBandForPower   = find(freq == (FreqUpperBound-f0)); % Final index for bandpower estimation: area will be estimated using approximation by the left
-
-BandPower      = f0*sum(Pxx(InitialIndexBandForPower:FinalIndexBandForPower));    % Power in the band specified
-
-PSDBandPowerOut = BandPower; % Defining the output

paradigma/tremor/feat_extraction/PSDEst.m

@@ -1,63 +0,0 @@
-% #########################################################################
-% Input
-% Conf - Configuration Strucutre
-%     .pwelchwin - (scalar) - double - number of samples - window for Welch periodogram
-%     .noverlap  - (scalar) - double - number of samples - ovserlap between windows
-%     .Fs        - (scalar) - double - sampling frequency
-%     .nfft      - (scalar) - double - number of samples for FFT evaluation
-%     .freqmax   - (scalar) - double - maximal frequency of interest
-%
-% Data           - (matrix) - double - IMU time series (Nsamples x 3 columns - x y z arrays)
-
-% Output
-% PSDMatrix - (matrix) - double - (Nfreq x 4 columns) - PSD coefficients for x, y 
-% and z and sum (fourth column: PSDx + PSDy + PSDz)  
-%
-% FreqVect  - array - double - Frequency array in Hz;
-%
-% FreqPeak  - scalar - double - Frequency (Hz) of the spectral peak
-% #########################################################################
-
-function [PSDMatrix,FreqVect,FreqPeak] = PSDEst(Conf,Data)
-
-% Get PSD estimating parameters from the configuration structure
-pwelchwin = Conf.pwelchwin;
-noverlap  = Conf.noverlap;
-Fs        = Conf.Fs;
-nfft      = Conf.nfft;
-freqmax   = Conf.freqmax;
-freqmin   = Conf.freqmin;
-
-% Get number of samples and number of signals to be manipulated
-[Nsamples,Nsignals] = size(Data);
-
-% Estimate the PSD coefficients (Pxx) for the frequencies (freqlong)  
-% hann - Hanning window
-[Pxx,freqlong] = pwelch(Data,hann(pwelchwin),noverlap,nfft,Fs);
-
-% Finding the index of the maximal frequency to be considered
-indcutfreq     = find(freqlong == freqmax);
-
-% Get the frequency vector up to be maximal frequency
-FreqVect       = freqlong(1:indcutfreq);
-
-% Store the PSD coefficients for each IMU signal [x y z] up the maximal
-% frequncy into a matrix (PSDMatrix)
-PSDMatrix = Pxx(1:indcutfreq,:);
-
-% Fourth column of PSDMatrix: sum of the PSD (PSDx + PSDy + PSDz)
-% This sum is used to work with a single orientation
-% independent spectrum
-PSDMatrix(:,Nsignals+1) = sum(PSDMatrix,2);
-
-% Get maximum PSD value in the range [freqmin freqmax]
-[CoeffPeak,IndPeak] = max(PSDMatrix(FreqVect>=freqmin,Nsignals+1));
-
-% Add the minimum frequency to obtain the correct dominant frequency
-FreqPeak = FreqVect(IndPeak) + freqmin;
-
-% Inspection Spectrum; 
-% figure; plot(FreqVect,PSDMatrix);
-% xlabel('Freq (Hz)'); ylabel('PSD');
-% set(gca,'fontsize',20,'fontweight','bold','LineWidth',2);
-% grid on; legend('PSDx','PSDy','PSDz','PSDsum');