Zoom out Search Issue

ManualsBrandsContents Manualsaudio & home theatreZoom in

IEEE SIGNAL PROCESSING MAGAZINE [40] MARCH 2015

microphone position .d

As in [27], the real factors are typically

applied which compensate for the amplitude change following

the

/r1 law, where r is the propagated distance. The second

gain (, )knH

,iHRTF

is a complex head-related transfer function

(HRTF) for the left or right ear, {, },i left right! respectively,

which depends on the DOA (, )kn

i with respect to the posi-

tion and look direction of the VL. Apart from creating a plausi-

ble feeling of being present in the actual classroom, the

user-defined spatial selectivity can be achieved with the third

gain

(, , ),dknB

IPLS

which enables the amplification or attenua-

tion of directional sounds emitted from d

IPLS

as desired. In prin-

ciple, any desired spatial selectivity function (, , )dknB can be

defined. For instance, a spatial spot can be defined at a teacher’s

desk or in front of a blackboard to assist the student in better

hearing the teacher’s voice. Such a gain function for a circular

spot centered around

spot

with a 1 m radius could be defined as

(, , )

;

dkn

otherwise

IPLS

(12)

where (, )ddknr

spot IPLS

=- and a controls the spatial selec-

tivity for the sources located outside the spot. In addition, the

gain

() [,]kQ 01

! applied to the diffuse component enables the

student to control the level of the ambient sound. The output

diffuse signals

(, )

,id

for the left and right headphone channel

are decorrelated such that the coherence between (, )knY

,leftd

and

(, )knY

d,right

corresponds to the target coherence in binaural

hearing [18], [29]. Finally, it should be noted that since the prop-

agation compensation and the spatial selectivity gains are typi-

cally real factors, the phase of the direct and diffuse components

are equal to those observed at the reference microphone. How-

ever, the complex HRTFs that dependent on the DOAs at the vir-

tual listening position ensure that the spatial cues are correct.

BINAURAL HEARING AIDS

Developments in acoustic signal processing and psychoacoustics

have lead to the advancement of digital hearing aids that were first

developed in the 1990s. The early devices included the unilateral

(i.e., single-ear) and bilateral hearing aids, where two independent

unilateral hearing aids are used for the left and right ears, respec-

tively. More recently binaural hearing aids, in which signals and

parameters can be exchanged between the left and right hearing

aid, have been brought to the market. Binaural hearing aids are

advantageous compared to unilateral and bilateral hearing aids as

they can further improve speech intelligibility in difficult listening

situations, improve the ability to localize sounds, and decrease lis-

tening fatigue. Besides dynamic range compression and feedback

cancelation, wind and ambient noise reduction, dereverberation

and directional filtering are important features of state-of-the-art

hearing aids.

Let us consider a situation in which we have one desired talker

in front and two interfering talkers at the right side of the hearing-

aids user, as illustrated in Figure 8. In such a situation, directional

filtering allows a hearing-aid user to perceive sounds arriving from

the front more clearly than the sounds from the sides. In addition,

one can aim at reducing the amount of diffuse sounds such that

the SDR increases.

While many state-of-the-art directional filtering techniques

for hearing aids are based on classical differential array pro-

cessing, some parametric spatial sound processing techniques

have been proposed. In [14], the left and right microphone sig-

nals were jointly analyzed in the time-frequency domain to

determine: 1) the interaural phase difference and interaural

level difference that strongly depend on the DOA of the direct

sound, and 2) the interaural coherence that measures the

degree of diffuseness. Based on these parameters, three gains

were computed related to the degree of diffuseness, signal-to-

interference ratio, and direction of the sound. Finally, real-val-

ued gains for the left and right microphones were determined

based on these gains to reduce reverberation and interfering

sounds. According to the authors of [14], the quality of the sig-

nal was good but the speech intelligibility improvement for a

single interfering talker was unsatisfactory. In [15], the authors

used two microphones at each side and adopted the DOA-based

geometric model. The DOAs were estimated at low frequencies

using the microphones at the left and respectively right side,

and at high frequencies using the intermicrophone level differ-

ences. Finally, the signal of a single microphone positioned at

the left and right, respectively, was modified based on the DOA

Desired

Source

Undesired

Sources

Spatial Analysis

Right

Processing

and

Synthesis

Processing

and

Synthesis

Left

Hearing-Aid

Signal

Right

Hearing-Aid

Signal

DOAs

Direct Sounds

Diffuse Sounds

[FIG8]

A general parametric spatial sound processing scheme for

binaural hearing aids.

THE WORLD’S NEWSSTAND

THE WORLD’S NEWSSTAND