function r = mirfeatures(x,varargin)
% f = mirfeatures(x) computes a large set of features from one or several
% audio files. x can be either the name of an audio file, or the
% 'Folder' keyword.
% mirfeatures(...,'Stat') returns the statistics of the features instead
% of the complete features themselves.
% mirfeatures(...,'Segment',t) segments the audio sequence at the
% temporal positions indicated in the array t (in s.), and analyzes
% each segment separately.
%(not available yet)
% mirfeatures(...,'Filterbank',nc) computes the analysis on each channel
% of a filterbank decomposition.
% Default value: nc = 5
% mirfeatures(...,'Frame',...)
% mirfeatures(...,'Normal')
% mirfeatures(...,'Sampling',s)
% miraudio options (Extract, ...)
[stat,nchan,segm,feat] = scanargin(varargin);
if isa(x,'miraudio') || isa(x,'mirdesign')
a = miraudio(x,'Normal'); % normalize with respect to RMS energy
% in order to consider timbre independently of
% energy
else
a = miraudio('Design','Normal');
end
if not(isempty(segm))
a = mirsegment(a,segm);
end
% DYNAMICS
% --------
r.dynamics.rms = mirrms(a,'Frame');
% Perceived dynamics: spectral slope?
% RHYTHM
% ------
%r.fluctuation = mirstruct; % not in Matlab Central version
%r.fluctuation.tmp.f = mirfluctuation(a,'Summary');
%r.fluctuation.peak = mirpeaks(r.fluctuation.tmp.f,'Total',1);%only one?
%r.fluctuation.centroid = mircentroid(r.fluctuation.tmp.f);
r.rhythm = mirstruct;
r.rhythm.tmp.onsets = mironsets(a);
%r.rhythm.eventdensity = ...
r.rhythm.tempo = mirtempo(r.rhythm.tmp.onsets,'Frame');
%r.rhythm.pulseclarity = mirpulseclarity(r.tmp.onsets,'Frame');
% Should use the second output of mirtempo.
attacks = mironsets(r.rhythm.tmp.onsets,'Attacks');
r.rhythm.attack.time = mirattacktime(attacks);
r.rhythm.attack.slope = mirattackslope(attacks);
% TIMBRE
% ------
f = mirframe(a,.05,.5);
r.spectral = mirstruct;
r.spectral.tmp.s = mirspectrum(f);
%pitch = mirpitch(a,'Frame',.05,.5);
r.spectral.centroid = mircentroid(r.spectral.tmp.s);
r.spectral.brightness = mirbrightness(r.spectral.tmp.s);
r.spectral.spread = mirspread(r.spectral.tmp.s);
r.spectral.skewness = mirskewness(r.spectral.tmp.s);
r.spectral.kurtosis = mirkurtosis(r.spectral.tmp.s);
r.spectral.rolloff95 = mirrolloff(r.spectral.tmp.s,95);
r.spectral.rolloff85 = mirrolloff(r.spectral.tmp.s,85);
r.spectral.spectentropy = mirentropy(r.spectral.tmp.s);
r.spectral.flatness = mirflatness(r.spectral.tmp.s);
r.spectral.roughness = mirroughness(r.spectral.tmp.s);
r.spectral.irregularity = mirregularity(r.spectral.tmp.s);
%r.spectral.inharmonicity = mirinharmonicity(r.spectral.tmp.s,'f0',pitch);
%r.spectral.mfcc = mirmfcc(r.spectral.tmp.s); %not in Matlab Central version
%r.spectral.dmfcc = mirmfcc(r.spectral.mfcc,'Delta');
%r.spectral.ddmfcc = mirmfcc(r.spectral.dmfcc,'Delta');
r.timbre.zerocross = mirzerocross(f);
r.timbre.lowenergy = mirlowenergy(f);
r.timbre.spectralflux = mirflux(f);
% PITCH
% -----
r.tonal = mirstruct;
r.tonal.tmp.chromagram = mirchromagram(a,'Frame','Wrap',0,'Pitch',0);
r.tonal.chromagram.peak=mirpeaks(r.tonal.tmp.chromagram,'Total',1);
r.tonal.chromagram.centroid=mircentroid(r.tonal.tmp.chromagram);
% TONALITY/HARMONY
% ----------------
keystrengths = mirkeystrength(r.tonal.tmp.chromagram);
[k r.tonal.keyclarity] = mirkey(keystrengths,'Total',1);
%r.tonal.keyclarity = k{2};
r.tonal.mode = mirmode(keystrengths);
r.tonal.hcdf = mirhcdf(r.tonal.tmp.chromagram);
if stat
r = mirstat(r);
% SHOULD COMPUTE STAT OF CURVES FROM FRAMED_DECOMPOSED HIGH FEATURES
end
if not(isa(x,'miraudio')) && not(isa(x,'mirdesign'))
r = mireval(r,x);
end
function [stat,nchan,segm,feat] = scanargin(v)
stat = 0;
nchan = 1;
segm = [];
feat = {};
i = 1;
while i <= length(v)
arg = v{i};
if ischar(arg) && strcmpi(arg,'Filterbank')
i = i+1;
if i <= length(v)
nchan = v{i};
else
nchan = 10;
end
elseif ischar(arg) && strcmpi(arg,'Stat')
i = i+1;
if i <= length(v)
stat = v{i};
else
stat = 1;
end
elseif ischar(arg) && strcmpi(arg,'Segment')
i = i+1;
if i <= length(v)
segm = v{i};
else
segm = 1;
end
else
feat{end+1} = arg;
end
i = i+1;
end