Zoom out Search Issue

IEEE SIGNAL PROCESSING MAGAZINE [73] MARCH 2015

that sound processing schemes based on the presentation of

selected acoustic features of speech signals were technically and

perceptually limited. As mentioned previously, most of those

schemes provided CI users with partial information primarily

about the two lowest speech formants (F1, F2) and the fundamen-

tal frequency (F0). While those schemes enabled many recipients

to understand speech adequately in favorable listening conditions,

performance was degraded by even moderate levels of background

noise. This was mainly because of the technical difficulty of esti-

mating parameters corresponding to the selected speech features

in real time when the signal-to-noise ratio is poor. The SMSP and

its successor schemes, Speak and ACE (as well as closely related

schemes provided by other CI companies), attempt to provide CI

users with information about salient aspects of the acoustic

[FIG4] (a) Waveform, (d) spectrogram, and (b)–(c) and (e)–(i) electrodograms of the sentence ”A boy fell from the window” from the

HINT corpus, uttered by a male speaker. All parameters are identical to those of Figure3.

–1

6,400

3,200

Frequency (Hz)Magnitude + Electrode

1,600

800

400

200

100

Relative Magnitude

Frequency (Hz)Magnitude + Electrode Relative Magnitude

–2

0 0.5 1

Waveform MP3000 EE

1.5 0 0.5 1 1.50 0.5 1 1.5

0 0.5 1 1.5 0 0.5 1 1.50 0.5 1 1.5

ACESpectrogram F0mod

HiRes120CIS-22ch FSP

Time (s) Time (s) Time (s)

(g) (h) (i)

(d) (e) (f)

(a) (b) (c)

“A Boy Fell from the Window”

a boy fell from the window

THE WORLD’S NEWSSTAND

THE WORLD’S NEWSSTAND