File Exchange

image thumbnail

Impulse response acoustic information calculator

version 1.5.4.0 (715 KB) by Christopher Hummersone
Calculate RT, DRR, Cte, and EDT for impulse response file

31 Downloads

Updated 05 May 2018

GitHub view license on GitHub

NOTE: this function is now available from the IoSR Matlab Toolbox as iosr.acoustics.irStats.
-------------------------
Calculate RT, DRR, Cte, and EDT for impulse response file
RT = IR_STATS(FILENAME) returns the reverberation time (to -60 dB)
using a method based on ISO 3382-1:2009. The function uses reverse
cumulative trapezoidal integration to estimate the decay curve, and a
linear least-square fit to estimate the slope between 0 dB and -60 dB.
Estimates are taken in octave bands and the overall figure is an
average of the 500 Hz and 1 kHz bands.
FILENAME should be the full path to an audio file or the name of an
audio file on the Matlab search path. The file can be of any format
supported by the AUDIOREAD function, and have any number of channels;
estimates (and plots) will be returned for each channel.
The function returns a 1xN vector of RTs, where N is the number of
channels in the audio file.
The function determines the direct sound as the peak of the squared
impulse response.
[RT,DRR] = IR_STATS(FILENAME) returns the direct-to-reverberant-ratio
DRR for the impulse; DRR is the same size as RT. This is calculated
in the following way:

DRR = 10 * log10( X(T0-C:T0+C)^2 / X(T0+C+1:end)^2 )

where X is the approximated integral of the impulse, T0 is the time of
the direct impulse, and C=2.5ms [1].

[RT,DRR,CTE] = IR_STATS(FILENAME) returns the early-to-late index CTE
for the impulse; CTE is the same size as RT. This is calculated in
the following way:

CTE = 10 * log10( X(T0-C:T0+TE)^2 / X(T0+TE+1:end)^2 )

where TE is 50 ms.

[RT,DRR,CTE,CFS] = IR_STATS(FILENAME) returns the octave-band centre
frequencies CFS used in the calculation of RT.

[RT,DRR,CTE,CFS,EDT] = IR_STATS(FILENAME) returns the early decay
time EDT, which is the same size as RT. The slope of the decay curve
is determined from the fit between 0 and -10 dB. The decay time is
calculated from the slope as the time required for a 60 dB decay.

... = IR_STATS(...,'PARAMETER',VALUE) allows numerous
parameters to be specified. These parameters are:

'graph' : {false} | true
Controls whether decay curves are plotted. Specifically, graphs
are plotted of the impulse response, decay curves, and linear
least-square fit for each octave band and channel of the audio
file. If the EDT output is specified, the EDT fit will also be
plotted.
'te' : {0.05} | scalar
Specifies the early time limit (in seconds).
'spec' : {'mean'} | 'full'
Determines the nature of RT and EDT outputs. With spec='mean'
(default) the reported RT and EDT are the mean of the 500 Hz
and 1 kHz bands. With spec='full', the function returns the
RT and EDT as calculated for each octave band returned in
CFS; RT and EDT have size [M N] where M=length(CFS).
'y_fit' : {[0 60]} | two-element vector
Specifies the decibel range over which the decay curve should
be evaluated. For example, 'y_fit' may be [-5 -25] or [-5 -35]
corresponding to the RT20 and RT30 respectively.
'correction' : {0.0025} | scalar
Specifies the correction parameter C (in seconds) given above
for DRR and CTE calculations. Values of up to 10 ms have been
suggested in the literature.

Octave-band filters are calculated according to ANSI S1.1-1986 and IEC
standards. Note that the OCTDSGN function recommends centre frequencies
fc in the range fs/200 < fc < fs/5.

The author would like to thank Feifei Xiong for his input on the
correction parameter.

References

[1] Zahorik, P., 2002: 'Direct-to-reverberant energy ratio
sensitivity', The Journal of the Acoustical Society of America,
112, 2110-2117.

See also AUDIOREAD, OCTDSGN.

Comments and Ratings (15)

jing zhang

@Mohsin can you post a link to the impulse response file that causes the error (or send the file to me directly).

I am having this error:

>> RT = iosr.acoustics.irStats('K:\EmoDB_distance\Release_RT_estimation_MatlabFileExchange\speech_file\TSP1min48.wav', 'graph', false, 'te', 0.05 , 'y_fit', [-5 -25], 'correction', 0.0025, 'spec', 'mean')
In an assignment A(I) = B, the number of elements in B and I must be the same.

Error in iosr.acoustics.irStats (line 191)
t0(n) = find(x(:,n).^2==max(x(:,n).^2)); % find direct impulse

I tried the following command

RT = irStats('K:\EmoDB_distance\Release_RT_estimation_MatlabFileExchange\speech_file\merged.wav', 'graph', true, 'te', 0.05 , 'spec', 'full', 'y_fit', [0 60], 'correction', 0.0025)

but cannot make it work. it says
'Undefined function 'irStats' for input arguments of type 'char'.' Could you please help me fix this issue?

@Razvan can you post a minimal working example so that I can try to reproduce and fix the bug? Please also post a link to the impulse response file that causes the error (or send the file to me directly).

I can not find calc_decay function

>> rt=irStats(filename);
Undefined function 'calc_decay' for input arguments of type 'double'.

Error in irStats (line 201)
[rt_temp(f,n),E_rt,fit_rt] = calc_decay(z(f,t0:end,n),options.y_fit,60,fs,cfs(f)); % estimate
RT

Thanks for the bug fix! Seems to work well now even with impulse responses having a low dynamic range.

@Earl, can you post a minimal working example so that I can try to reproduce and fix the bug?

I often get:

Assignment has more non-singleton rhs dimensions than non-singleton subscripts

Error in IR_stats (line 186)
[rt_temp(f,n),E_rt,fit_rt] = calc_decay(z(f,t0:end,n),options.y_fit,60,fs);

Hi Augusto. Yes you can. Try:

Rt60 = ir_stats(filename,'spec','full');

Chris

Hi there,

Is it possible with this .m file to determine de RT60 for the separate octave bands? I was trying to do so but I couldn't figure it out. Tried to pass a filtered IR generated with the audiowrite function but it didn't work out.

Hi Clay. I need a bit more information in order to help. Firstly, did you download the Octave toolbox per the requirements above (and add the files to your MATLAB search path)?

Clay Pipkin

An error comes up:

Error in IR_stats (line 164)
[b(f,:),a(f,:)] = octdsgn(cfs(f),fs,N);

Any thoughts?

Updates

1.5.4.0

Migrated to GitHub.

1.5.4.0

Made decay time calculation more robust to input files with a low dynamic range (thanks to Earl Vickers for the bug report).

1.5.3.0

Added more input checking.

1.5.2.0

Updated description.

1.5.2.0

A number of improvements and clarifications to the help. Some minor refactoring to improve the clarity of code.

1.5.1.0

Added error trap for dynamic range in IR.

1.5.0.0

Changed inputs to accept parameter/value pairs, including new 'correction' parameter. Changed default correction.

1.4.0.0

Corrected t0 estimation and H1 line.

1.3.0.0

Made a number of tweaks: added more comments, added EDT output, and made slight modification to RT calculation so that y range can be specified and is [0 -60] by default.

1.2.0.0

Minor tweak to file and updated screenshot.

1.1.0.0

Minor to tweak to handling out-of-range te values.

MATLAB Release Compatibility
Created with R2013a
Compatible with any release
Platform Compatibility
Windows macOS Linux
Acknowledgements

Inspired by: octave