Beamforming using a spherical microphone array - Nils Peters
Transcript of Beamforming using a spherical microphone array - Nils Peters
Beamforming using a spherical microphone array based on legacy microphone characteristics
Nils Peters•+
[email protected] W. Schmeder•
•Center for New Music and Audio Technologies (CNMAT), UC Berkeley+International Computer Science Institute (ICSI)
Berkeley, California, US
Acknowledgments
References[1] A. Farina, A. Capra, L. Chiesi, and L. Scopece. A spherical microphone
array for synthesizing virtual directive microphones in live broadcasting and in post production. In Proc. of the AES 40th International Conference, Tokyo, Japan, 2010.
[2] R. Avizienis, A. Freed, T. Suzuki, and D. Wessel. Scalable connectivity processor for computer music performance systems. In International Computer Music Conference, pages 523–526, 2000.
The SAM-arrayAbstract
• To measure and match other legacy recording microphones, dummy-heads and surround mic setups
• Modeling nonlinear behaviors of microphones• Subjective evaluation (e.g., ABx listening test)• Real-time implementation
Thanks to Meyer Sound for using their anechoic chamber and technical support. Nils Peters is supported by the German Academic Exchange Service (DAAD). Thanks to Lorenzo Chiesi for clarifying details of [1].
We present an approach for projecting frequency-dependent directivity pattern of classic recording microphones onto steerable beams created with a spherical microphone array. In an anechoic chamber, the spatial and timbral characteristics of a classic recording microphone and the characteristics of a novel 120-channel spherical microphone array is measured. Using a least-square matching approach, these impulse responses were used to derive the set of filters that synthesize the desired legacy recording microphone characteristic from the spherical microphone array.
Motivation
Future work
The SphericAl Microphone array is a co-development of CNMAT and Meyer Sound and consists of 144 microphones.
Sphere: Sensors: 120 x DPA 4080 cardioid capsules Radius: 6.9 cm Theoretical spatial resolution: > 9 Rigid body: Icosahedron Radius: 4.5 cm
Outriggers:24 x DPA 4060 omnidirectional capsules 3 sensors per outriggerspherical layout with three diameters
Onboard electronics:Mic-amps and AD converter 96kHz / 24 bit Wordclock syncAudio Network I/O [2] FPGA
1. Recording and post-production processes could be simplified through flexible beamforming technology offered by spherical microphone arrays.
2. Tonmeister often reject spherical microphone arrays and prefer classic (legacy) recording microphones
3. This might be due to a preference for distinct spatial and timbral characteristics of different microphone types and brands.
Least-Square ApproachTo synthesize the measured recording microphone, we adopted the least-square approach [1] to our needs.The idea is to create the set of FIR filters (h) that filters the microphone signals of the spherical microphone array so that the sum of all microphone signals xn creates a signal y which approximates the desired target directivity pattern q.
SAM Anonymus recording microphone
[Ck]D×M
WM×M
βk
...Matrix of 576 x 120 SAM frequency responses measured from 36 x 16 directions, bin k ...Frequency response of the anonymus recording microphone, measured from 36 x 16 directions, bin k
[Hk]M×V =[Ck]∗M×D ·WM×M · [Qk]D · e−jπk
[Ck]∗M×D ·WM×M · [Ck]D×M + βk · [I]M×M
[Qk]D...Weighting matrix to equalize sampling grid (10 degrees in azimuth and elevation) ...Frequency-dependent regularization parameter, for frequency bin k
A recording microphoneWe measured the frequency response of a mid-price cardioid recording microphone in an anechoic chamber.
100 250 500 1000 2500 5000 10000
0
1
2
3
4
5
6
7
8
9
Frequency [Hz]
Ord
er
of
sph
erica
l ha
rmo
nic
s
Le
vel c
on
trib
utio
n in
[d
B]
−30
−25
−20
−15
−10
−5
0
10010 20
100
1020
20
0
20
125 Hz
100
1020
100
102020
0
20
250 Hz
010
2010
010
2020
0
20
500 Hz
010
2010
010
2020
0
20
1000 Hz
100 10 20
20
0
2020
0
20
2000 Hz
10010 20
100
102040
20
0
4000 Hz
10010 20
20
0
2020
10
0
10
8000 Hz
100
1020
5051020
10
0
10
16000 Hz
The spherical harmonics decomposition reveals contribution of higher orders harmonics, not just zero and first order.
Results
Directivity plot of the measured recording microphone
Frequency dependent contribution of spherical harmonics to the overall microphone directivity
102 103 1040
2
4
6
8
10
Mag
nitu
de[d
B]
Frequency [Hz]
Ideal mic array syhthesizing ideal cardioidIdeal mic array synthesizing measured reference microphoneMeasured mic array syhthesizing ideal cardioidMeasured mic array synthesizing measured reference microphoneSimulated gain error ± 3dBSimulated alignment error ± 1 sample @ 44.1 kHz
Conclusion• Good results for low and mid frequencies• Non-optimal results above 3kHz• Challenges: phase alignment of SAM measurements, noise
102 103 104100
80
60
40
20
0Magnitude and Phase of all Mics
Mag
nitu
de[d
B]
102 103 1048
6
4
2
0
2 x 104 Phase
Phas
e [d
eg]
Frequency [Hz]125 Hz, Bin 4 250 Hz, Bin 6
100
1020
100
10
20
10
0
10
x
Target
y 100
1020
100
10
20
10
0
10
x
Result
y
Frequency [Hz]
505 1015
100
1020
10
0
10
x
125 Hz, Bin 4
y 1001020
1001020
40
20
0
20
x
250 Hz, Bin 6
y0 1020
200
20
40
20
0
20
x
500 Hz, Bin 12
y 1001020
200
20
40
20
0
20
x
1000 Hz, Bin 22
y
1001020
200
20
40
20
0
20
x
2000 Hz, Bin 44
y100
102020
020
40
20
0
x
4000 Hz, Bin 86
y100
102020
020
40
20
0
20
x
8000 Hz, Bin 172
y 5 0 5 10
50
510
5
0
5
x
16000 Hz, Bin 342
y
Error Analysis
Simulated SAM Measured SAM
h1(t)
h2(t)
h120 (t)
+.:
x1(t)
x2(t)
x120(t)
.:
y(t)
102 103 10480
60
40
20
0
20
40Magnitude and Phase of all Mics
Mag
nitu
de[d
B]
102 103 104
6
4
2
0
x 104 Phase
Phas
e [d
eg]
Frequency [Hz]125 Hz, Bin 4 250 Hz, Bin 6
Frequency [Hz]
505 1015
100
1020
10
0
10
x
125 Hz, Bin 4
y 1001020
1001020
40
20
0
20
x
250 Hz, Bin 6
y 1001020
200
20
40
20
0
20
x
500 Hz, Bin 12
y 1001020
200
20
40
20
0
20
x
1000 Hz, Bin 22
y
1001020
200
20
40
20
0
20
x
2000 Hz, Bin 44
y100
102020
020
40
20
0
x
4000 Hz, Bin 86
y100
10 2010
010
2020100
10
x
8000 Hz, Bin 172
y0 5 10
50
510
5
0
5
x
16000 Hz, Bin 342
y
100
1020
100
10
20
10
0
10
x
Target
y 100
1020
100
10
20
10
0
10
x
Result
y