mirattackslope(orig,varargin)

function varargout = mirattackslope(orig,varargin)
%   a = mirattackslope(x) estimates the average slope of each note attack. 
%   Optional arguments:
%   a = mirattackslope(x,m) specifies a method for slope computation.
%       Possible values:
%           m = 'Diff': ratio between the magnitude difference at the 
%               beginning and the ending of the attack period, and the
%               corresponding time difference.
%           m = 'Gauss': average of the slope, weighted by a gaussian
%               curve that emphasizes values at the middle of the attack
%               period. (similar to Peeters 2004).
%       mirattackslope(...,'Contrast',c) specifies the 'Contrast' parameter
%           used in mironsets for event detection through peak picking.
%           Same default value as in mironsets.
%
% Peeters. G. (2004). A large set of audio features for sound description
% (similarity and classification) in the CUIDADO project. version 1.0

        meth.type = 'String';
        meth.choice = {'Diff','Gauss'};
        meth.default = 'Diff';
    option.meth = meth;
    
        cthr.key = 'Contrast';
        cthr.type = 'Integer';
        cthr.default = NaN;
    option.cthr = cthr;
    
specif.option = option;

varargout = mirfunction(@mirattackslope,orig,varargin,nargout,specif,@init,@main);


function [o type] = init(x,option)
o = mironsets(x,'Attack','Contrast',option.cthr);
type = mirtype(x);


function sl = main(o,option,postoption)
if iscell(o)
    o = o{1};
end
po = get(o,'PeakPos');
pa = get(o,'AttackPos');
pou = get(o,'PeakPosUnit');
pau = get(o,'AttackPosUnit');
sr = get(o,'Sampling');
d = get(o,'Data');
sl = mircompute(@algo,po,pa,pou,pau,d,option.meth,sr);
fp = mircompute(@frampose,pau,pou);
sl = mirscalar(o,'Data',sl,'FramePos',fp,'Title','Attack Slope');
sl = {sl,o};


function fp = frampose(pa,po)
pa = sort(pa{1});
po = sort(po{1});
fp = [pa';po'];


function sl = algo(po,pa,pou,pau,d,meth,sr)
pa = sort(pa{1});
po = sort(po{1});
pau = sort(pau{1});
pou = sort(pou{1});
sl = zeros(1,length(pa));
for i = 1:length(pa)
    switch meth
        case 'Diff'
            sl(i) = (d(po(i))-d(pa(i)))/(pou(i)-pau(i));
        case 'Gauss'
            l = po(i)-pa(i);
            h = ceil(l/2);
            gauss = exp(-(1-h:l-h).^2/(l/4)^2);
            dat = diff(d(pa(i):po(i))).*gauss';
            sl(i) = mean(dat)*sr;
    end
end