Zoom out Search Issue

IEEE SIGNAL PROCESSING MAGAZINE [109] MARCH 2015
as critical as the use of nonindividualized HRTFs [13]. Thus,
equalization using individual HPTFs is strongly recommended.
Another difficulty in carrying out accurate headphone equaliza-
tion is the variability of the HPTFs with repositioning. The effect of
repositioning of headphones is lower at low frequencies but dis-
plays high standard deviations up to 10 dB at high frequencies
[37]. Kulkarni et al. [37] observed that equalization based on a sin-
gle measurement may become worse than no equalization at all.
The positional dependency has no specific solution and its effect
can only be reduced by taking the average of a number of trials as
a representative HPTF [37]. Thus, to create a convincing immer-
sive sound environment, use of individualized HRTFs and individ-
ualized equalization is entailed, which may not be viable all the
time. To reduce the dependency on individualized equalization,
Sunder et al. [33] designed a Type-2 equalization technique for the
playback through frontal projection headphone, which is indepen-
dent of the headphone-ear coupling. Unlike the conventional
equalization technique, Type-2 equalization compensates only for
the distortion due to the emitter, thereby preserving the individual
pinna cues due to frontal projection.
The other type of equalization is the “decoupled” equalization
technique, and it is the most commonly used method of equaliza-
tion for rendering music. In this technique, the binaural record-
ing [(BIR) or HRTFs] as well as the headphone are equalized
using a reference sound field (REF) (e.g., FF, DF, etc.) [36]. If the
REF of the recording environment is well known and reproduced
reliably, this method of equalization can result in a very natural
perception of sound similar to the nondecoupled equalization
technique. This method of equalization is mainly carried out to
make the binaural recordings compatible with stereophonic (con-
ventional microphone) recordings in terms of timbral quality.
If the recording is binaural, then a reference field equalized
binaural recording (BIR/REF) achieves a sound quality equivalent
to a conventional microphone recording. When the equalized
recording is played from a reference field equalized headphone
(HPTF/REF), the perceived timbre of the spatial sound would be as
natural as the original binaural recording. Individualized binaural
recordings are thus necessary to experience the true immersive-
ness of sound without any timbral coloration and spatial degrada-
tion. Note that for rendering conventional stereo recorded music,
it is sufficient to carry out just the headphone equalization using
an appropriate reference field. Some of the commonly used refer-
ence fields are:
■ Free-field (FF) equalization: With the aim to replicate the ear
signals produced by frontal loudspeakers, the target response of
FF equalization is the HRTF of frontal incidence. Hammershoi
et al. proposed an FF equalization curve, which has additional
high frequency energy above 3 kHz to approximate listening to
stereo loudspeakers in the FF [4]. A FF equalized headphone
can reproduce a frontal sound with natural sound quality but
colors the sound that originates from other directions. More-
over, it is important to note that there are large interindividual
variations in the FF equalization filters [38].
■ Diffuse-field (DF) equalization: In this case, the target
response for equalization is the DF response, i.e., the average
of the HRTFs of all measured directions in horizontal plane.
The interindividual variations are reduced drastically due to
the averaging effect [38]. Thus, the DF target response can be
achieved universally over a great number of individuals.
Møller [35] identified certain headphones which are already
DF equalized and recommended such type of headphones for
stereo listening.
■ Other target responses: A typical listening room is not
completely diffuse but it can be considered somewhere
between a FF and a DF. Møller [38] illustrated other alterna-
tive target responses which are partially diffuse by applying
unequal weighting to different directions within ± 45° azi-
muth and elevation. Other researchers also modified the DF
equalization filters with the help of certain parametric fil-
ters and found that the subjects generally preferred the tar-
get response with a 3 kHz peak lower in amplitude than in
the DF response for both music and speech [4]. Recent
[TABLE 4] EQUALIZATION TECHNIQUES FOR DIFFERENT PLAYBACK MODES (BINAURAL, STEREOPHONY).
MODE OF
EQUALIZATION AIM
TYPES OF
EQUALIZATION AND
TARGET RESPONSE CHARACTERISTICS
NONDECOUPLED
(BINAURAL)
SPECTRUM AT
EARDRUM IS THE
INDIVIDUAL HRTF
FEATURES
CONVENTIONAL
EQUALIZATION (FLAT
TARGET RESPONSE)
■ FOR CONVENTIONAL HEADPHONES. THE SPECTRUM AT
THE EARDRUM HAS INDIVIDUAL FEATURES (IF INDIVIDUALIZED HRTF IS USED)
■ DEPENDENT ON THE INDIVIDUAL’S UNIQUE PINNA FEATURES
TYPE-2 EQUALIZATION
[33]
■ FOR FRONTAL PROJECTION HEADPHONES. THE SPECTRUM AT EARDRUM
AUTOMATICALLY MODELS THE INDIVIDUAL PINNA SPECTRAL CUES
■ REMOVES ONLY THE DISTORTION DUE TO THE HEADPHONE EMITTER
■ INDEPENDENT OF THE IDIOSYNCRATIC FEATURES OF THE EAR
DECOUPLED
(BINAURAL,
STEREOPHONY)
EMULATE
THE MOST NATURAL
REPRODUCTION
CLOSER TO THE
PERCEPTION IN A
REFERENCE FIELD
FF EQUALIZATION [38] ■ TARGET RESPONSE IS THE FF RESPONSE CORRESPONDING TO
THE FRONTAL INCIDENCE
DF EQUALIZATION [38] ■ TARGET RESPONSE IS THE DF RESPONSE
■ LESSER INTERINDIVIDUAL VARIABILITY
DF TARGET RESPONSE
BASED ON MØLLER [38]
■ TARGET RESPONSE BASED ON AVERAGE OF HRTFS BETWEEN ± 45 DEGREES
AZIMUTH AND ELEVATION WITH UNEQUAL WEIGHTING
DF TARGET RESPONSE
BASED ON LORHO [4]
■ REDUCED A 3-KHZ PEAK FROM ABOUT 12 DB TO 3 DB OF DF RESPONSE
RR_G AND RR1_G [4] ■ RR_G: BASED ON THE IMPULSE RESPONSE OF HARMAN
REFERENCE LISTENING ROOM
■ RR1_G HAS LESSER BASS AND TREBLE
Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page
q
q
M
M
q
q
M
M
q
M
THE WORLD’S NEWSSTAND
®
Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page
q
q
M
M
q
q
M
M
q
M
THE WORLD’S NEWSSTAND
®