Zoom out Search Issue

ManualsBrandsContents Manualsaudio & home theatreZoom in

IEEE SIGNAL PROCESSING MAGAZINE [70] MARCH 2015

at which pitch changes are perceived, around 300 Hz. Further-

more, temporal envelope fluctuations are not always accurately

coded by current sound processing strategies.

Currently, an increasing number of people are being implanted

bilaterally, especially children. Also, due to relaxed implantation

criteria, an increasing number of people can make use of bimodal

stimulation. These CI recipients have residual hearing in the non-

implanted ear, which can be aided with an acoustic hearing instru-

ment. Listeners with bilateral CIs or using bimodal stimulation

can potentially perceive interaural time differences (ITDs). There-

fore another topic of intensive research is binaural hearing and the

preservation of binaural cues in applications with bilateral and

bimodal devices. ITDs, the difference in arrival time between the

ears, are important binaural cues for NH listeners to localize

sound sources and to separate multiple sound sources such as

speech and noise. The latter is called binaural unmasking. ITDs

range from 0 μs for sounds in front to around 700 μs for sounds

from the side of the head. NH listeners can use ongoing temporal

cues that are present in both the fine structure and the envelope of

sound signals, and temporal cues in the onset of signals.

BASIC STIMULATION STRATEGIES

Historically, the first main types of stimulation strategies can be clas-

sified as feature extraction strategies. In such strategies, estimates of

F0 and formants F1 and F2 of speech signals are calculated in real

time. Formants are peaks in the spectral envelope corresponding to

resonances of the vocal tract. Formants are used by the auditory sys-

tem to identify sounds such as vowels. The formant information is

used predominantly to stimulate channels corresponding to F1 and

F2. The F0 is used to control the pulse rate. The outcomes in speech

understanding of these schemes are, on average, lower than those of

more recent schemes, and therefore they are not normally used any

more in commercial processors [2[,[4].

A simple strategy, widely used in CI signal processing, is

continuous interleaved sampling (CIS); see Figure 2 and [7].

CIS is based on a running spectral analysis of the preprocessed

digital input sound signal performed by a bank of band-pass fil-

ters or a fast Fourier transform (FFT). The filter bank has an

overall bandwidth from approximately 100 to 8,000 Hz, and

the number of filters usually equals the number of stimulation

channels at the electrode array-neuron interface. The filters

have partially overlapping frequency responses and bandwidths

that generally become broader with increasing frequency. Each

filter is assigned to (at least) one intracochlear electrode fol-

lowing the frequency-place tonotopic organization of the

cochlea. Although the correspondence of signal frequencies

and filter bank outputs to depth of electrode insertion follows

the tonotopy, the signal is not necessarily delivered to

[FIG2] A block diagram of all monaural strategies discussed in this article. Common elements are shown in (a) and (c), while strategy-

specific elements are shown in (b).

Envelope

Detection

Filter Bank

Front-End

Processing

Microphones

F0mod

eTone

Modulation

Enhancement

Channel

Selection

F0 Feature

Extraction

MP3000

Channel

Selection

Masking Model

Onset

Enhancement

Channel

Selection

Channel

Selection

ACE

CIS

FSP

Temporal Feature

Enhancement

HiRes120

Current Steering

Mapping

Electrical

Stimulation

(a)

(b)

(c)

THE WORLD’S NEWSSTAND

THE WORLD’S NEWSSTAND