Say hello. Hello!-4 Clap your knees –clap,clap,clap-2 Clap your hands-clap, clap, clap-2
The Hand Clap as an Impulse Source for Measuring Room ...
Transcript of The Hand Clap as an Impulse Source for Measuring Room ...
The Hand Clap as an Impulse Source forMeasuring Room Acoustics
Prem Seetharaman, Stephen P. Tarzia
Northwestern University
April 11, 2012
1 / 27
Why?
The Upshot
Claps are easy to produce.Reliable acoustic measurements from claps allows amateurs todiagnose acoustic problems quickly and easily.
2 / 27
Background
Room acousticsThere are three qualities which determine a room’s sound.
1. Space
2. Warmth
3. Clarity
3 / 27
Background
Room acousticsThere are three qualities which determine a room’s sound.
1. Space - Reverberation time
2. Warmth - Frequency response
3. Clarity - Frequency decay
4 / 27
Background
Impulse response
Definition: how a room reacts to a short, loud sound.
Ideal impulse response
1. Very short
2. Covers a wide frequency range
3. Consistent
5 / 27
Background
Good impulse sources
1. Balloon pops
2. Starter pistol shots
3. Firecrackers, etc.
6 / 27
Motivation
An amateur acoustician’s ideal workflow
1. Clap a few times in a room.
2. Have a program quickly respond with analysis.
3. Diagnose and improve room based on analysis.
7 / 27
Hand Claps
What’s in a clap?
Advantages:
1. Easily produced
Disadvantages:
1. Low energy
2. Long duration
3. Inconsistent
4. Mediocre frequency coverage
9 / 27
Setup
Recording
Clap from the stage, record from the center with a Zoom H4Nrecorder.
Datasets
1. Pick-Staiger Concert Hall (Northwestern Univ.): 19 claps.
2. Lutkin Concert Hall (Northwestern Univ.): 12 claps.
3. Makeshift home recording studio: 6 claps.
Ground truthA professional acoustic survey performed by Northwestern onPick-Staiger Concert Hall.
11 / 27
Segment claps
Thresholding
Input: a series of claps.
1. Clap onset: power is ten times the background level.
2. Clap end: power returns to twice the background level.
Output: individual claps.
12 / 27
Reverberation time
DefinitionRT60: time required for a sound level to decay by 60 dB.Problem: 60 dB above background level = jet engine − a dial tone.Solution: Extrapolate the rest via a line-fitting approximation.
13 / 27
Reverberation time
Problems with claps
Low energy: end of clap (the ”knee” point) is difficult to find.Finding the ”knee”:
1. Too early: linear fit doesn’t use enough data.
2. Too late: linear fit is too shallow.
3. Getting it just right:
3.1 Try all possible knee points after the first 100 ms, andminimize error of the resultant fit line.
0 0.2 0.4 0.6 0.8 1−45
−40
−35
−30
−25
−20
−15
Time (s)
Ener
gy (
dB
)
Example fit line for a clap
← Direct sound
↓ Knee point
Linear decay region
Direct sound
Fit line
Background
14 / 27
Reverberation time
0 0.2 0.4 0.6 0.8 1−45
−40
−35
−30
−25
−20
−15
Time (s)
Ener
gy (
dB
)
Example fit line for a clap
← Direct sound
↓ Knee point
Linear decay region
Direct sound
Fit line
Background
15 / 27
Reverberation time
ResultsHighly encouraging. For Pick-Staiger Concert Hall:
1. Mean: 1.74 seconds.
2. Standard deviation: .07 seconds.
Ground truth comparison
Overall reverberation results unavailable, but close to middle bandfrequency results from the acoustic survey.
16 / 27
Frequency decay
DefinitionRT60 of a particular frequency.Process:
1. Split recording up into frequencies using a spectrogrammethod.
2. Pass each signal to our RT60 method.
17 / 27
Frequency decay
Problems with claps
Lacks low freqency information.Balloons also lack low frequencies, but claps provide about anoctave less than balloons do.
18 / 27
Frequency decay
Results: Pick-Staiger Concert Hall
0
1
2
3
4
5
6
2
8
8
1
25
1
77
2
50
3
54
5
00
7
07
10
00
14
14
20
00
28
28
40
00
56
57
80
00
11
31
4
Frequency (Hz)
RT
60 r
ever
ber
atio
n t
ime
(s)
balloon pop
19 / 27
Frequency decay
Results: Lutkin Concert Hall
0
1
2
3
4
5
6
2
8
8
1
25
1
77
2
50
3
54
5
00
7
07
10
00
14
14
20
00
28
28
40
00
56
57
80
00
11
31
4
Frequency (Hz)
RT
60
rev
erb
erat
ion
tim
e (s
)
20 / 27
Frequency decay
Results: Makeshift home recording studio
0
1
2
3
4
5
6
2
8
8
1
25
1
77
2
50
3
54
5
00
7
07
10
00
14
14
20
00
28
28
40
00
56
57
80
00
11
31
4
Frequency (Hz)
RT
60
rev
erb
erat
ion
tim
e (s
)
21 / 27
Frequency response
DefinitionRatio of reverberant and direct sound spectra.Process:
1. Split signal up into frequency bins
2. Divide reverberant sound spectrum by direct sound spectrum.
22 / 27
Frequency response
Results: Frequency response for Pick-Staiger Concert Hall
−5
0
5
3
1
4
4
6
2
8
8
1
25
1
77
2
50
3
54
5
00
7
07
10
00
14
14
20
00
28
28
40
00
56
57
80
00
11
31
4
16
00
0
Frequency (Hz)
Frequency responsesE
ner
gy (
dB
)
23 / 27
Clap consistency
Not all claps are created equal?
Claps can vary in intensity and frequency spread.We can account for this using the direct sound spectrum.
24 / 27
Clap consistency
Results: Direct sound clap spectra for Pick-Staiger concert Hall
−65
−60
−55
−50
−45
−40
−35
−30
3
1
4
4
6
2
8
8
1
25
1
77
2
50
3
54
5
00
7
07
10
00
14
14
20
00
28
28
40
00
56
57
80
00
11
31
4
16
00
0
Frequency (Hz)
Direct sound spectraN
orm
aliz
ed e
ner
gy
(d
B)
25 / 27
Conclusions
Reverberation TimeReliable.
Frequency Decay
Reliable for frequencies > 300 Hz.
Frequency Response
Reliable for middle frequencies (mid and treble response).
26 / 27