Zoom out Search Issue

ManualsBrandsContents Manualsaudio & home theatreZoom in

100

IEEE SIGNAL PROCESSING MAGAZINE [90] MARCH 2015

radiation from the rear of the driver, although the directivity

results can be less accurate with frequency [7].

ARRAYS OF DIRECTIONAL SOURCES

If multiple directional loudspeakers are available, then it becomes

possible to create multiple zones of sound. Multizone reproduc-

tion requires a large number of monopole loudspeakers. The use

of directional sources allows the production of multizone fields

using significantly fewer loudspeaker units. In effect, a large num-

ber of drivers are grouped into a small number of physical devices

to allow the creation of complex sound fields.

It has been shown that an array of

LNth order sources

operating in free-field conditions has a spatial Nyquist fre-

quency of approximately

N2 times that of the same geometry

monopole array [31]. Results better than free-field can be

achieved in a reverberant room by using the techniques dis-

cussed in [32]. In this case, the directional sources are able to

exploit room reflections to provide directions of arrival other

than those directly from the sources. The use of

L HOSs, each

of which can produce up to order N responses, can produce a

similar accuracy of a reconstructed field to ()L N21+ mono-

pole loudspeakers in the 2-D case, and ()L N 1

+ loudspeakers

in the 3-D case. For example, Figure 8 shows the sound field

reproduction error achieved using a circular array of five

higher-order loudspeakers in comparison with an array of 45

monopole sources. For a virtual source angle of 72° (the

desired source position is equal to the first loudspeaker posi-

tion), the error is similar to that produced by the monopole

sources. At the angle of 36° (the desired source halfway

between two loudspeakers), the error is about 10 dB higher

than the monopole case but still reasonably accurate, particu-

larly at low frequencies. Reproduction has been achieved over a

1-m diameter using only five loudspeaker units with room

dereverberation. The simulation is limited to 2-kHz bandwidth

for computational complexity reasons. The worst-case repro-

duction error will be below

10- dB up to around 3 kHz. The

bandwidth and reproduction radius of accurate reproduction

can be extended by using more sources and higher orders, cre-

ating sufficient space for multiple listeners listening to inde-

pendent sound fields.

The use of HOSs can be viewed as an optimization problem

with a constraint on the total number of loudspeaker units in the

room. The only way to improve reproduction in such a case is to

add capability to the existing loudspeakers. HOSs offer a practical

approach to providing the control of the high-spatial-dimension

sound fields that are required for creating multiple personal sound

zones. For example, the reproduction of sound in

Q zones of

radius ,r

up to a spatial frequency ,k

max

using L HOSs requires a

maximum order per source of

(.)

..N

Qk r 05

max 0

(19)

For 8 kHz reproduction over regions of radius 0.2 m, the order is

N 10= for L 10= sources and N 6= for L 15= sources. Such

numbers are achievable in moderate- to large-sized rooms.

SUMMARY AND FUTURE OPPORTUNITIES

In this article, we presented, according to our involvement and

insights, the audio processing and loudspeaker design aspects

that support the goal of establishing personal sound zones. The

problems that have been explored include multizone sound con-

trol, wave-domain active room compensation, and directional

loudspeaker design, which allow for sound control over spatial

regions. A high-performance personal audio system would likely

address many of these aspects in its design. In sound field

control, interference mitigation and room compensation robust

to changes and uncertainty in the acoustic environment remain

as challenging problems. Yet future opportunities exist in

1) higher-order surround sound using an array of directional

sources and wave-domain active room compensation to perform

multizone sound control in reverberant enclosures and 2) per-

sonal audio devices using multiple sensors to establish personal

sound zones by efficiently canceling crosstalk and using distrib-

uted beamforming.

AUTHORS

Terence Betlehem (Terence.Betlehem@callaghaninnovation.govt.

nz) received the B.S., B.E., and Ph.D. degrees in telecommunica-

tions engineering from the Australian National University (ANU)

in 1998, 2000, and 2005 respectively. From 2005 to 2006, he was a

research fellow at the ANU Research School of Information Sci-

ences and Engineering and a visiting researcher at National Infor-

mation and Communication Technology Australia working in the

areas of spatial signal processing and wireless channel modeling.

Since 2007, he has worked at Callaghan Innovation (formerly

Industrial Research Limited) in Lower Hutt, New Zealand, in the

areas of spatial audio and wireless communications, where he is

currently a senior research engineer. His research interests

0 500 1,000 1,500 2,000

−50

−40

−30

−20

−10

Frequency (Hz)

Least Square Error (dB)

45 Omni

Sources

5 HOS, 36°

5 HOS, 72°

[FIG8] The least squares error of reproduction as a function of

frequency for an array of five fourth-order sources at 36˚ exactly

between a pair of loudspeakers (dashed) and 72˚ coinciding with

a loudspeaker (dashed), and a circular array of 45 omnidirectional

line sources (unbroken) in a 2-D rectangular room of dimensions

6.4 # 5 m and with wall reflection coefficients of 0.7.

THE WORLD’S NEWSSTAND

THE WORLD’S NEWSSTAND

__________________________