calc_diff_metrics.m

我认为很不错的语音处理的ｍａｔｌａｂ源代码

function [s]=calc_diff_metrics(s, diff_chan, ratio_metrics, round_kind, round_digits)
% % calc_diff_metrics: Calculates the differences in metrics between pairs of channels: Impulsive Noise Meter
% %
% % Syntax:  
% %
% % [s]=calc_diff_metrics(s, diff_chan);
% %
% % *****************************************************************
% %
% % Description:
% %
% % This program takes the output from the Impulsive_Noise_Meter and
% % calculates the differences in metrics between pairs of channels: 
% % Impulsive Noise Meter.
% % 
% % This program is specifically for analyzing hearing protection for
% % impulsive noise.
% % 
% % *****************************************************************
% %
% % Input Variables
% %
% % s is the data structure created using the Impulsive_Noise_Meter.
% %
% % diff_chan=[1,2];    % A paired column vector of channel numbers.
% %                     % 
% %                     % The odd and even elements of the vector are 
% %                     % paired for calculating differences
% %                     % in metrics across channels.  
% %                     % 
% %                     % default is diff_chan=(1:2)';
% % 
% % ratio_metrics=[3, 4, 5, 20];    % is an array of indices of metrics for
% %                                 % the diff_chans array that are
% %                                 % calculated as ratios instead of
% %                                 % differences
% %                         
% % round_kind=1;           % Array of values one element for the rta array
% %                         % and one element for each varargin array
% %                         % (see example)
% %                         % 1 round to specified number of significant
% %                         % digits
% %                         %
% %                         % 0 round to specified digits place
% %                         %
% %                         % default is round_kind=1;
% %
% % round_digits=3;         % Array of values one element for the rta array
% %                         % and one element for each varargin array
% %                         % (see example)% Type of rounding depends on round_kind
% %                         %
% %                         % if round_kind==1 number of significant digits
% %                         % if round_kind==0 specified digits place
% %                         %
% %                         % default is round_digits=3;
% % 
% % *****************************************************************
% % 
% % Output Variables
% % 
% % s is the data structure created using the Impulsive_Noise_Meter.
% % 
% % *****************************************************************
% 
% Example='s';
%
% % This is an example using shock tube data! The data compares two
% % data acquisition rates.
%
% % An example which outputs the mean of the metrics.
%
% load shock_tube;
% stat_to_get=1;
% fileout
% [bb_table, bb2]=make_table_compare_systems(s, stat_to_get, fileout);
%
%
% % An example which outputs all of the metrics.
%
% load shock_tube;
% stat_to_get=[1:7];
% fileout
% [bb_table, bb2]=make_table_compare_systems(s, stat_to_get, fileout);
%
%
%
% % *****************************************************************
% % 
% % Subprograms
% % 
% % 
% % 
% % List of Dependent Subprograms for 
% % calc_diff_metrics
% % 
% % 
% % Program Name   Author   FEX ID#
% % 1) genHyper		Ben Barrowes		6218	
% % 2) LMSloc		Alexandros Leontitsis		801	
% % 3) m_round		
% % 4) pow10_round		
% % 5) sd_round		
% % 6) t_alpha		
% % 7) t_confidence_interval		
% % 8) t_icpbf		
% % 
% % 
% % *****************************************************************
% % 
% % Written by Edward L. Zechmann
% % 
% %     date 29 August      2008
% % 
% % modified 30 August      2008    Updated comments.
% % 
% % modified 17 September   2008    Updated comments.
% %
% % modified  7 January     2009    Updated comments.
% %
% % modified 18 January     2009    Updated to include rounding. 
% %
% %
% % *****************************************************************
% %
% % Please feel free to modify this code.
% %
% % See Also:  make_summary_impls_stats_table, Impulsive_Noise_Meter, Continuous_Sound_and_Vibrations_Analysis
% %


if nargin < 1 || isempty(s) || ~iscell(s)
    load('shock_tube');
    warning('Loading and analyzing the default shock tube data.');
end

if nargin < 2 || isempty(diff_chan) || ~isnumeric(diff_chan)
    diff_chan=(1:2)';
end

if nargin < 3 || isempty(ratio_metrics) || ~isnumeric(ratio_metrics)
    ratio_metrics=[3, 4, 5, 20];
end

if nargin < 5 || isempty(round_kind) || ~isnumeric(round_kind)
    round_kind=1;
end

if nargin < 5 || isempty(round_digits) || ~isnumeric(round_digits)
    round_digits=3;
end

[num_files, num_vars]=size(s);
num_channels_array=zeros(num_files,1);




% Determine the size of the concatenated metrics table
for e1=1:num_files;    % Data files

    num_channels=[];

    for e3=1:num_vars; % Number of Variables (Number of Data Acquisition Systems)

        if ~isempty(s{e1,e3})

            [num_metrics, num_channels, num_stats2]=size(s{e1,e3}.stats_of_metrics);

            if num_channels >= 1

                num_channels_array(e1)=max([num_channels, num_channels_array(e1)]);

            end
        end

    end

end

num_diff_chan=floor(length(diff_chan)/2);

% Determine the length of the one-dimensional rounding arrays.
num_kinds=length(round_kind);
num_digits=length(round_digits);

% Fill in the concatenated metrics table
% Add row headings and column headings

for e1=1:num_files;    % Data files

    for e3=1:num_vars; % Number of Variables (Number of Data Acquisition Systems)

        if ~isempty(s{e1,e3})
            
            [num_metrics, num_channels, num_stats2]=size(s{e1,e3}.stats_of_metrics);

            if num_channels >= 1

                s{e1,e3}.diff_chan=diff_chan;
                % calculate the differences, then the statistics then
                % save them to the structure s.
                for e2=1:num_diff_chan; % Channels

                    for e4=1:num_metrics; % Data Metrics

                        buf1=s{e1,e3}.metrics{diff_chan(2*e2-1), 1}(:, e4);
                        buf2=s{e1,e3}.metrics{diff_chan(2*e2), 1}(:, e4);

                        [m1_buf1, n1_buf1]=size(buf1);
                        [m1_buf2, n1_buf2]=size(buf2);

                        m1_buf3=min([m1_buf1 m1_buf2]);
                        n1_buf3=min([n1_buf1 n1_buf2]);

                        
                        if num_kinds >= (e4)
                            rk=round_kind(e4);
                        else
                            rk=1;
                        end
                        
                        if num_digits >= e4
                            rd=round_digits(e4);
                        else
                            rd=3;
                        end
                        
                        
                        % Determine whether to compute a difference or
                        % ratio.  Calculate a ratio for metrics  that
                        % are a member of ratio_metrics.
                        if ismember(e4, ratio_metrics) 
                            rk=1;
                            rd=3;
                            buf3=abs(buf1(1:m1_buf3,1:n1_buf3))./abs(buf2(1:m1_buf3,1:n1_buf3));
                        else
                            buf3=abs(buf1(1:m1_buf3,1:n1_buf3))-abs(buf2(1:m1_buf3,1:n1_buf3));
                        end
                        
                        s{e1,e3}.diff_metrics{e2,1}(:,e4)=m_round(buf3, rk, rd);

                        % Calculate the Arithmetic Mean
                        amean_avg=mean(buf3);
                        
                        % Calculate the Robust Mean
                        bmean_avg=LMSloc(buf3);

                        % Calculate Standard Deviation
                        stdrt=std(buf3);             
                        
                        % Calculate the 95% Confidence Interval  of
                        % the standard error of the
                        % t-distribution with a two-sided test.
                        [ci_int]=t_confidence_interval(buf3, 0.95);                  
                        
                        % Calculate the Median
                        medianrt=median(buf3);
                        
                        % Calculate the Median Index
                        [mbuf ix]=min(abs(buf3-medianrt));
                        
                        % Calculate the Minimum
                        minrt=min(buf3);
                        
                        % Calculate the Maximum
                        maxrt=max(buf3);

                        % Set the metric values to the cell array
                        s{e1,e3}.diff_stats_of_metrics(e4, e2, 1)=m_round(amean_avg, rk, rd);
                        s{e1,e3}.diff_stats_of_metrics(e4, e2, 2)=m_round(bmean_avg, rk, rd);
                        s{e1,e3}.diff_stats_of_metrics(e4, e2, 3)=m_round(stdrt, 1, 3);
                        s{e1,e3}.diff_stats_of_metrics(e4, e2, 4)=m_round(ci_int, rk, rd);
                        s{e1,e3}.diff_stats_of_metrics(e4, e2, 5)=ix;
                        s{e1,e3}.diff_stats_of_metrics(e4, e2, 6)=m_round(medianrt, rk, rd);
                        s{e1,e3}.diff_stats_of_metrics(e4, e2, 7)=m_round(minrt, rk, rd);
                        s{e1,e3}.diff_stats_of_metrics(e4, e2, 8)=m_round(maxrt, rk, rd);

                    end
                end

            end
        end

    end

end