Sound strength parameter G determined from uncalibrated room impulse responses measurements of concert halls using a set of window functions and lengths

Published: 20-01-2021| Version 3 | DOI: 10.17632/rhgkb8wfxj.3
Ernesto Accolti,
Fernando di Sciascio


The sound strength factor G computed using two hundred and eighty seven time windows in six frequency bands from two hundred and seventy seven uncalibrated room impulse responses (RIRs) corresponding to seven concert halls is reported. The RIRs were collected from the website of the Concert Hall Research Group of the Acoustical Society of America. The raw, processed, and compared data are provided as coma delimited, mat files and figures as well as the codes implementing the method. The maximum error across halls for mid frequencies using Flat-Top window of 30 ms was 1 dB. Figures p1, p2, p3 and p4 show the summary of the data for the 1/1 octave band centered at 125 Hz, the low frequencies, the mid frequencies, and the high frequencies, respectively. The summary data are the mean error, maximum error and correlation across halls for the average G of each hall using seven time window functions and forty-one time window lengths for each window function. The whole data can be accessed in the folder DAT_files. The data is divided in 6 groups. In the first group, the strength factor G is estimated for the six 1/1 octave frequency bands from 125 Hz to 4 kHz. Each file contains a table of six columns corresponding to the frequency bands and a number of rows corresponding to the source-receiver combinations. Each file is labelled G_R_W_L.dat. R codes the concert hall: BM: Baltimore. Joseph Meyerhoff Hall, 30 BS: Boston. Symphony Hall, 42 BK: Buffalo. Kleinhans Music Hall, 72 PA: Philadelphia. Academy Of Music, 36 WM: Worcester. Massachusetts. Mechanics Hall, 25 CS: Cleveland. Severance Hall, 36 WK: Washington. D.C. JFK Center, 36 W codes the window function: TW: Taylor, c=32 FT: Flattop, c=21 RW: Rectangular, c=100 HN: Hann, c=46 HM: Hamming, c=48 BW: Blackman, c=37 BH: Blackman Harris, c=31 L codes the minimum required time for a 5 dB rejection criterion. The codes are the following: 00, 01, …, 40. The window length can be calculated as Δt=2 c/100 L, where the value of c is given in the previous list. A group (b) of two hundred and eighty seven files contains the average G for each of the seven halls in each row. This second group is coded as G_b_W_L.dat. A group (c) comprises the average for low, mid , and high frequencies coded as G_c_low.dat, G_c_mid.dat, and G_c_hig.dat. The columns of each file correspond to the halls and the rows to the windows. Each row correspond to a window length and each consecutive group of forty-one rows correspond to a window function. (e.g. the first forty-one correspond to a Taylor window with L of 0 ms, 1 ms, 2 ms, …, 40 ms, respectively). Groups (d), (e), and (f), are the data reported in figures p1-p4 obtained by comparison with results of G reported in the literature. All this data can be generated using the Matlab code result_figures.m and the Mat files. In turn, the mat files can be generated using the code calculate_Num_Den.m and the other auxiliary codes provided in the Matlab_code folder.