IMPLEMENTATION OF THE DEVELOPMENT OF A FILTERING ALGORITHM TO IMPROVE THE SYSTEM OF HEARING IN HEARING IMPAIRED WITH COCHLEAR IMPLANT

IMPLEMENTATION OF THE
DEVELOPMENT OF A FILTERING
ALGORITHM TO IMPROVE THE SYSTEM
OF HEARING IN HEARING IMPAIRED
WITH COCHLEAR IMPLANT
Salaheddine Derouiche#1 and Bachir Djedou#2
#

Laboratory of study, research in Instrumentation and Communication of
Annaba.
University Badji Mokhtar, BP 12, Annaba, 23000 Algeria.
1

der-salah@hotmail.fr , 2djedoubachir@yahoo.fr

ABSTRACT
In this paper, we present the implemented denoising section in the coding strategy of cochlear
implants, the technique used is the technique of wavelet bionic BWT (Bionic Wavelet
Transform). We have implemented the algorithm for denoising Raise the speech signal by the
hybrid method BWT in the FPGA (Field Programmable Gate Array), Xilinx (Virtex5
XC5VLX110T). In our study, we considered the following: at the beginning, we present how to
demonstrate features of this technique. We present an algorithm implementation we proposed,
we present simulation results and the performance of this technique in terms of improvement of
the SNR (Signal to Noise Ratio). The proposed implementations are realized in VHDL (Very
high speed integrated circuits Hardware Description Language). Different algorithms for
speech processing, including CIS (Continuous Interleaved Sampling) have been implemented
the strategy in this processor and tested successfully.

KEYWORDS
BWT implementation, CIS Strategy, cochlear implant, FPGA, speech processing & FIR digital
filter.

1. INTRODUCTION
J. Yao and Y.T. Zhang introduced in their research work [1, 2], Bionic wavelet transform (BWT)
as adaptive wavelet transform. It is specially built to model the human auditory system [3]. The
BWT is based on the model Giguere-Woodland auditory system [4], a model incorporating
feedback electro-active auditory canal, middle ear and cochlea [3]. The term'' Bionic'' means that
a mechanism guided by biologically active [5]. The BWT differs from the conventional Wavelet
Transform (WT) in that the time-frequency resolution by the BWT completed, can be adaptively
adjusted not only by the frequency variations of the signal, but also by the amplitude
instantaneous signal and the derivative of order 1 [5, 1]. It is through the mother wavelet, the
wavelet transform classical adaptive whereas the BWT is the active control mechanism in the
human auditory model that adjusts according to the mother wavelet signal analysis [5 ]. Primarily,
the basic idea of the BWT is inspired by the fact that we want to make the shell of the mother
Sundarapandian et al. (Eds) : ICAITA, SAI, SEAS, CDKP, CMCA-2013
pp. 65–79, 2013. © CS & IT-CSCP 2013

DOI : 10.5121/csit.2013.3807

Computer Science & Information Technology (CS & IT)

66

wavelet time-varying according to the characteristics of the signal and this by introducing a
function T (a,τ ) variable in time, in the expression of the mother wavelet ϕ according to the
equation 1 [5].
ϕ (t ) =



1
T (a,τ ) a

t 
 ⋅ exp jω 0 t

 T (a,τ ) 

(

~
ϕ


)

(1)

with ω0 denotes the fundamental frequency of the mother wavelet.
In practice ω0 is equal to 15165.4 for the human auditory system [2]. In the formula (1), the role
of the first factor T (a,τ ) multiplying a is to ensure that energy remains the same for each
mother wavelet and the second factor T (a,τ ) is to adjust the casing of the mother wavelet, ϕ (t )
without adjusting the center frequency [3]. Thus the adaptive nature of the BWT is captured by
the time varying factor, T (a,τ ) representing the scaling of the quality factor of the filter banks of
the cochlear, Qeq in each scale over time. Incorporating this directly to the scale factor of the
Morlet wavelet, we obtain:
XBWT(a,τ ) =

 t −τ 

a 

1
~∗  t −τ  − jω0 
∫ x(t) ⋅ϕ  T(a,τ ) ⋅ a  ⋅ e 


T(a,τ ) a



(2)

dt

~
where ϕ is the amplitude envelope of the mother wavelet Morlet which is given by:
  t
  T0







~
ϕ (t ) = exp  − 


2






(3)

with T0 is the initial time length of the carrier.
0.5
Real part
Imag part

0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
-0.5
-4

-3

-2

-1

0

1

2

3

4

Figure 1. The Morlet wavelet

The discretization of the scale variable a is performed by using the logarithmically spaced across
the predefined frequency range desired, such that the center frequency for each scale, is given by
equation (4 ) [3].
ˆ
D app

dtl

(ω )

With m=0,1,2,…

(4)

For this implementation, based on the original work of Yao and Zhang cochlear implant for the''
coding'' [9], the coefficients in 22 scales, m=7,…,28 , are calculated using the integration of
digital continuous wavelet transform [3]. These 22 scales are logarithmically spaced center
frequencies of 225 Hz to 5300 Hz. The adjustment factor T (a,τ ) for each scale and time is
calculated using the equation of adjustment [5, 3]:

67

T (a,τ ) =

1

 
Cs

1 − G
 ⋅ 1 + G ⋅ ∂ X

1
2
BWT (a, τ ) 


∂t
C s + X BWT (a,τ )  




(5)

Where G1 and G2 are active gain factor discussed in [5, 3]. Cs is a constant representing the
effects of nonlinear saturation in the model cochlea [2, 3].
The values of G1 , G2 and Cs , dependent on the properties of the target signal and are respectively
equal to 0.87, 45, 0.8 according to reference [2]. In practice, the partial derivative in equation (5)
is approximated using the first difference of the previous points of BWT at this scale [3]. From
equation (2), we can see that the adaptation factor T (a,τ ) has an influence on the duration of the
amplitude envelope of the wavelet but admits no influence on the frequency of the complex
exponential associated. Thus, the BWT can be seen as a mechanism for adapting the time support
of the wavelet by quality factor Qeq of the filter model and this cochlear corresponding to each
scale [3]. It has been demonstrated [11] that the wavelet coefficients obtained by application of
the BWT, X BWT (a,τ ) may beings deducted by multiplying the wavelet coefficients, X CWT (a,τ ) by
a constant K (a,τ ) which is a function of the adaptation factor T (a,τ ) . For the Morlet wavelet, it is
adaptive scaling factor expressed by equation 6 .
K (a,τ ) =

π
C

T0
1 + T 2 (a,τ )

(6)

Where C is a normalization constant calculated from the integral of the square of the mother
wavelet [3].
The relationship thus obtained between the X BWT (a,τ ) and X BWT (a,τ ) is given by equation (7).
X BWT (a,τ ) = K(a,τ ) ⋅ X CWT (a,τ )

(7)

This representation reduces to a computationally efficient method for calculating wavelet
coefficients directly from the bionic wavelet coefficients obtained by the application of the
Continuous Wavelet Transform (CWT). This without resorting to numerical integration of
equation (2) at each scale [3]. There exist various distinctions between the discretized CWT used
for the BWT transform and wavelet packet based on the notion of filter bank and using an
orthogonal wavelet as Daubechies family. These differences include the perfect reconstruction in
the use of wavelet packet transform; however the discretized CWT is an approximation whose
accuracy is dependent on the number and placement of selected frequency bands. Another
difference lies in the fact that the Morlet mother wavelet consists of a single frequency with a
support decaying time exponentially; however support frequency orthonormal wavelet families
used for the dyadic wavelet transform, DWT and the wavelet packet transform, WPT, covers a
wider band [3]. Thus the Morlet wavelet is more concentrated frequentially throughout each level,
which allowed a direct adaptation of temporal support with minimal impact on the support
frequency, central mechanism of adaptation of the BWT [3]. The rest of this paper is organized as
follows. In Section 2 and 3, we describe the Algorithm of noise reduction by bionic wavelet
transform. In Section 4, we describe the main design of bionic wavelet transform in VHDL. In
Section 5 and 6, we show the simulation of the new implementation of BWT denoising algorithm.
And conclude the paper by discussing open problems and research challenges in Section 7.


68

2. ALGORITHM OF NOISE REDUCTION BY BIONIC WAVELET TRANSFORM
Technical noise reduction by bionic wavelet transform, BWT, is summarized by the block
diagram is given in Figure 2 [5, 1].

Figure 2. The block diagram of the algorithm for speech enhancement by BWT

The continuous wavelet coefficients are calculated using discrete convolutions in each of the 22
scales using continuous wavelet transform, CWT:
1

X CWT (a,τ ) = x(t ), ϕ a ,τ (t ) =

a

∫ x(t ) ⋅ ϕ

∗

 t −τ 

dt
 a 

(8)

The equations (5) and (6) are used to calculate the factor K that represents the adjustment of the
time support of the wavelet coefficients obtained by application of the continuous wavelet
transform, CWT.
In their approach, Michael T. et al [3] used the safe method [15] for calculating the threshold:


n





k =1



ˆ
ˆ
λ = arg min σ 2 n + ∑ [min (x (k ) , λ 2 ) − 2σ 2 I ( x (k ) ≤ λ )]
ˆ
0 ≤ λ ≤σ

2 ⋅log n

(9)

Where I denotes the indicator function.

( )

(

)

λ
S doux d l ,k = signe(d l ,k ) d l ,k − λ avec α + = sup(α ,0)

(10)

+

with λ denotes the threshold used.
With this method, they used a threshold independent of the level and soft thresholding function
whose expression is given by equation (10). Another approach to speech enhancement introduced
by A. Sumithra MG et al [10] comprises applying to the noisy signal the Discrete Wavelet
Transform (DWT) instead of the continuous wavelet transform discretized (CWT). The wavelet
coefficients are then obtained multiplied by the factor K(a,n) for the adaptation time. These
coefficients are then thresholded using the soft thresholding function. The coefficients are then
1
thresholded multiplied by a factor
and the inverse discrete wavelet transform is then
K (a , n )
applied to obtain the enhanced signal. This approach is summarized by the algorithm given in
Figure 3[10].

69


Figure 3. The block diagram of the proposed approach

2.1. The Denoising by Wavelet
The choice of the wavelet [13] can be tricky, but it is very important because it determines the
quality of the results. This choice can be guided by:
A good compromise between time / frequency resolution,
Its mathematical properties: vanishing moments, regularity, size medium ...
A shape close to a pattern that you want to highlight in the signal without knowing the exact
scale. Each type of mother wavelet has its own characteristics (shape, regularity, vanishing
moments, time-frequency extension, etc). We therefore choose a particular analyzing wavelet
depending on what you want to highlight in the signal. Orthogonal wavelets are wavelets first
appeared in the work of Meyer and Mallat. A number of commonly used orthogonal families such
as Daubechies wavelets. The non-orthogonal wavelet families are called wavelet
biorthorthogonales. In fact, the prefix "bi" recalls that two wavelet bases are used, one is for the
decomposition and reconstruction [14].
2.1.1. The Haar Wavelet

The first example of orthogonal wavelet is the Haar wavelet. It is fairly standard [14]. The mother
wavelet is given by:
−1 si 0 ≤ x ≤ 0.5
1 si 0.5 < x ≤1

ψ(x) = 

(11)

2.1.2. The Daubechies Wavelet

These wavelets, especially suitable for multiresolution analysis, are not symmetrical and have a
compact support [14]. Figure.5 represents the wavelet functions for different vanishing moments.
There are other wavelets belonging to Daubechies system, the most important are: wavelets and
wavelet symlets coiflets.


70

Figure 4. Examples wavelet (a) Haar wavelet, (b) Daubechies wavelet of order 2, (c) Daubechies wavelet of
order 4, (d) Daubechies wavelet of order 6.

3. THE CODING STRATEGY OF MULTI CHANNEL COCHLEAR IMPLANT
Single electrode cochlear implant (CI) are failed mainly due to stimulate only a particular place in
the cochlea due to the single electrode used. Thus single electrode CI can only provide very
limited frequency information, since they use only one electrode and perform simple spectral
analysis. To better exploit the place/frequency mechanism found in the peripheral auditory
system, multi-channel CI were developed. Multi channel implants provide electrical stimulation
at multiple sites in the cochlea using an array of electrodes. An electrode array is used so that
different auditory nerve fibers can be stimulated at different places in the cochlea, thereby
exploiting the place mechanism for coding frequencies. Different electrodes are stimulated
depending on the frequency of the signal. Electrodes near the base of the cochlea are stimulated
with high-frequency signals, while electrodes near the apex are stimulated with low-frequency
signals.
The waveform based speech processing algorithms try to present some type of waveform (in
analog or pulsatile form) derived from the speech signal by filtering into different frequency
bands, while the feature-extraction based speech processing algorithms are try to present some
type of spectral features, such as formants, derived using feature extraction algorithms. Hybrid
algorithms presents the utilizing both algorithms. A brief coverage of these speech processing
algorithms is given in Figure 5.Here the speech processor design using FPGA architecture system
software comprises of two important modules, namely Programmable Speech Processing modules
and Speech Data Encoding modules as shown in figure 6.
The Programmability of the speech processing system design described herein provides the
means to develop and test 8 Channel Continuous Interleaved Sampled (CIS) speech processing
algorithms. It provides flexibility and programmability according to patient’s active electrodes.
By using the impedance telemetry and clinical programming software, the audiologist identifies
the active electrodes and sends this information to the Speech Processing module via Speech Data
Encoding module of Xilinx, device FPGA Virtex5 XC5VLX110T card. [16]
The Band pass filters are configured so that the cutoff frequencies are adjusted depending on the
number of active electrodes as shown in Table 1 and observe the block diagram of CIS algorithm
in the figure 7.
The input signals are digitally filtered into 8 band-pass filters using Hamming window finite
impulse response (FIR) filtering.. The input samples are stored in the circular buffer managed by

71


Xilinx Virtex5 FPGA. Each input acoustic sample is fed to a bank of band-pass channels. Each
channel includes the stages of band-pass filtering, envelope detection, compression and
modulation.

Figure 5. Classification coding strategies used in cochlear implants. [17-18].

Figure 6. Functional Modules of the Speech Processing Design

Figure 7. Block Diagram of CIS Algorithm

The temporal envelope in each channel is extracted with full-wave rectifier followed by 32nd
order low pass Hamming window FIR filter.


72

Table 1. Cut-off frequencies of Channels

The low-pass filters were designed to smooth the amplitude estimates in time, and were set to a
cutoff frequency of 200Hz to allow maximal transmission of temporal envelope cues while
avoiding aliasing when a relatively low carrier rates are used.

3.1. Realization of FIR filters
To test our filter, we use the method previously seen with the direct digital synthesis, The
following program generates a signal whose frequency increases from 5 kHz to 15 kHz and 20
kHz, all for a period of 3.14 µS. The program is written in VHDL and synthesized using the
Xilinx ISE14.2 tool. Here is the diagram obtained for our filter Figure 8 :

Figure 8. complete diagram of FIR filter in FPGA

73


Using software ScopeFIR determining the coefficients of the low pass filter for a sampling
frequency of 50 KHz, a band at -3 dB from 5 KHz and a 20 dB stopband 15 KHz, 6 coefficients
of 9 bits. Then we get:
•
•

•

h0=h5=0.01171875D
h1=h4=0.1875D
h2=h3=0.35546875D

and here is the output signal S[7..0] (Figure 9):

Figure 9.The output of the filtered signal

4. THE MAIN DESIGN OF BIONIC WAVELET TRANSFORM IN VHDL
Figure 10 The following program, generates a signal whose frequency increases from 5 kHz, then
to 15 kHz to 20 kHz, all during a period of 3,14 µs. The program is set over four parties, the
command block, the memory block, the operator block & comparator block.

Figure 10. Complete diagram of BWT denoising in FPGA


74

5. SIMULATION
In this section we will present our approach to speech enhancement by thresholding both in the
field of wavelet packet in the area of bionic wavelet transform (BWT).

5.1.The wavelet packet tree used
This method is to first apply the wavelet packet transform discrete DWPT followed by the
cancellation of some terminal nodes of the tree of wavelet packet thresholding then obtained to
remaining terminal nodes using function hard thresholding modified µ − loi given by (11). [12]

r si r > λ


r


S (r , λ ) =   (1 + µ ) λ − 1




 sign(r ) si
λ  
 

µ

 

 

(12)
r <λ

The elimination of these subbands is based on the fact that the speech signal is voiced consisting
areas, areas unvoiced and quiet zones. Voiced sounds are quasi-periodic in the time domain and
harmonically structured. In the frequency domain, they are usually located in the bands below 1
KHz. However, the energy of unvoiced sounds is often concentrated at high frequencies
(≥3KHz).If we want to distinguish between voiced sounds and unvoiced sounds, we should
benefit from the information contained in these frequency bands where the sound is voiced or
unvoiced dominant. It is well known that the power of the speech signal is contained around the
first form.
The statistical results for several vowels of male and female voices, indicate approximately the
frequency of the first formant does not exceed 1KHz and is not less than 100KHz. In addition, the
fundamental frequency of a normal voice, is between 80 and 500 Hz. From here you can cancel
the sub-bands which are less than 80 Hz and the remaining terminal nodes of the tree of wavelet
packet obtained by applying the DWPT are subject to the thresholding operation by using the
hard thresholding modified µ − loi . Speech signals used in the evaluation of this technique and
other techniques are 20 sentences uttered by two Arabic speakers (ten sentences by a female voice
and the other by a male voice). The sampling frequency is 16 KHz and the treatment is carried out
frame by frame. Each frame contains 512 samples with an overlap of 256 samples. The mother
wavelet used in this method is the Daubechies wavelet, which has 40 dB greater perceptual
property. The tree of wavelet packet employee is given in Figure 11.

75


Figure 11. The tree of wavelet packet used

Nodes are set to zero in this tree are the nodes (7,0) and (9,4) because they occupy subbands that
are lower than 80 Hz. And these nodes are considered noise and are those things used to reestimate the noise level in the neighboring nodes: before the cancellation of the node (7,0), it is
used to estimate the noise level in the node (7,1). The same applies to the node (9,4) is employed
to estimate the noise level in the node (9,5) , prior to its cancellation.

5.2. Simulation and implementation in FPGA
A test program was run on FPGA based speech processor with core frequency at 326 MHz as the
main processing unit, the ADC sampling rate at 64 kHz and the encoding module of FPGA
formatting and transmitting two data frames via the RF coil. A 128th order FIR program
containing 8 band-pass filters runs on Xilinx Virtex5 FPGA processing unit. The input signal
FPGA to the speech processor is received from the microphone.
Since the sampling rate is fixed at 64 KHz, we get 64000 samples for every second (i.e. 64
samples for every 1ms). These samples are fed to the filter bank, containing 8 band pass filters
with frequencies ranging from 200Hz to 6600Hz. Rectified signals are generated from each filter,
which are then fed to the low pass filter with cutoff frequency of 200Hz for envelope outputs.
These 8 envelope signals are compressed using power law compression. These compressed
signals of eight band-pass filters are transmitted to the encoding module in a base to apex order
as shown in fig 10. This process is continuous and repeated for every 1.15ms. The FPGA based
speech processor using Continuous Interleaved Sampling ( CIS) strategy [6,7,8] has been run and
tested with a laboratory model implant module.
and here is the output signal (Figure 12), (With the coefficients A and B of high and low
frequency of transformation).


76

Figure 12. The output of the BWT module.

6. RESULTS AND DISCUSSION
Figures 13 (a), (b), (c) and (d) show that there is virtually no loss of information when we put the
nodes (7,0) and (9,4) to zero .

Figure 13. Perfect reconstruction after cancellation of nodes (7.0) and (9.4)

77


Table 2 gives the values of SNRf (Signal to Noise Ratio after enhancement) obtained by applying
the first method of denoising by thresholding and this for different values of (Signal to Noise
Ratio before enhancement) SNRi (-5, 0, 5, 10 and 15dB) and two types of broadband noise
(white noise) and narrowband (Volvo car noise). For both male and female voices, SNRf values
in this table 2 are average values.

Figure 14. (a) Case of Volvo noise, (b) Cases of white noise

Table 2 and Figure 14 show that the first method improves the signal to noise ratio (SNRi ≤
SNRf).They also show that the results obtained in the case of female voices are better than those
obtained in the case of the male voice.
Table 2. Average SNR calculated by applying the first thresholding method using modified thresholding
function µ − loi .

SNRf (dB)
SNRi (dB)

male voice
Volvo
white
noise
noise

female voice
Volvo
white
noise
noise

-5

0.36

-2.46

0.50

-2.37

0

4.83

2.37

5.27

2.59

5

8.44

6.60

9.52

7.21

10

10.67

9.64

12.75

11.19

15

11.69

11.29

14.43

13.56

7. CONCLUSIONS AND FUTURE DIRECTIONS
In this paper, we proposed an approach to implementing one of the most used algorithms in
cochlear implant via a development type Virtex5 FPGA card equipped with a chip of type
XC5VLX110T. The main goal of this work is the enhancement of speech signal for the
rehabilitation of deafness by cochlear implant.


78

The advantage of this technique is that the mother wavelet analysis allows a shell variable with
time and adapts according to the characteristics of the speech signal analysis, which improves the
performance of the thresholding technique.
Although this success is encouraging, there is still much to learn about strategies electrical
stimulation in the speech enhancement on cochlear implants, and many questions to be answered.
In addition, reducing the number of channels with the intelligibility of the signal.

REFERENCES
[1]
[2]
[3]
[4]
[5]

[6]
[7]

[8]
[9]
[10]

[11]

[12]

[13]
[14]

[15]
[16]
[17]

Yao J., ‘‘An active model for otoacoustic emissions and its application to time-frequency signal
processing,’’ PhD., The Chinese University of Hong Kong, Hong Kong, 2001.
Yao J., Zhang, Y. T., ‘‘Bionic Wavelet Transform: a new time-frequency method based on auditory
model,’’ IEEE Trans. Biomed. Eng., 48 (8), 856-863, 2001.
Michael T. Johnson, Xiaolong Yuan, Yao Ren, ‘‘Speech signal enhancement through adaptive
wavelet thresholding,’’ ScienceDirect, Speech communication 49 (2007) 123-133, 2007.
Giguere C. Woodland P.C., ‘‘A computational model of the auditory periphery for speech and hearing
research,’’ J. Acoust. Soc. Amer. 95(1), 331-342, 1995.
Xiaolong Yuan, B.S.E.E., ‘‘Auditory Model-based Bionic Wavelet Transform for Speech
Enhancement,’’ PhD thesis, MARQUETTE UNIVERSITY, Speech and Signal Processing Lab
Milwaukee, Wisconsin, May 2003.
C. Philipos LOIZOU "Speech Processing in Vocoder-Centric Cochlear Implants" Advances in
Otorhinolaryngology Basel, Karger, vol. 64, p. 109-143,(2006).
Fan-Gang Zeng, Stephen Rebscher William HARRISON, Xiaoan SUN & FENG Haihong: "Cochlear
Implants: System Design, Integration, and Evaluation" IEEE Reviews in Biomedical Engineering,
vol. 1, p. 115-142, (2008).
Ay, S., Zeng, F. -G. and Sheu, B. (1997). “Hearing with bionic ears: “Speech processing strategies
for cochlear implant devices“ , IEEE Circuits & Devices Magazine, May 1997, 18-23.
L. Breiman, ‘‘Better Subset Regression using the Non-negative Garrote,’’ Technometrics, vol. 37, pp.
327-384, 1995.
A. Sumithra M G, B. Thanuskodi K, C. Anitha M R, ‘‘Modified Time Adaptive Based Approach for
Enhancing Speech from Adverse Noisy Environments,’’ DSP Journal of ICGST, Volume 9, Issue 1:
33-40. June 2009.
Yao J., Zhang Y.T., ‘‘The application of bionic wavelet transform to speech signal processing in
cochlear implants using neural network simulations’’. IEEE Trans. Biomed. Eng. 49 (11), 1299-1309,
2002.
Talbi M., Salhi L., Chérif Adnane, ‘‘Spectral entropy Employment in Speech Enhancement based on
Wavelet Packet,’’ International Journal of Computer and Information Science and Engineering,
pp.136-143, 2007.
L’équipe technique Gresilog Traitement du signal, Bibliothèque ondelettes
version1.0.
http://www.Lis.inpg.fr/mustig/doc_html/ondlet/ondlet.htm.
Morgan BRISHOUAL, ‘‘Reconstruction des données, application à la dosimétrie des
Radiotéléphones’’, Thèse de doctorat, Electronique, L’Institut National des Sciences Appliquées de
Rennes, Octobre 2001.
D. Donho, I.M. Johnstone, ‘‘Adapting to unknown smoothness by wavelet shrinkage,’’ Journal of the
American statistical association, (90) 41, pp. 1200-1224, 1995.
Site : Xilinx Platform FPGAs. “Available from”: http://www.xilinx.com.
C. Philipos LOIZOU: "Mimicking the Human Ear"IEEE Signal Processing Magazine, p. 101-130,
(1998).

79


Authors
Salaheddine DEROUICHE received his master in electrical engineering from UBMA
University, ANNABA, Algeria in 2010, and searcher in LERICA laboratory at Annaba,
Algeria since 2011. Currently, he is an assistant professor at Badji-Mokhtar University,
Annaba, Algeria since 2011. His interests are in speech processing.
Djedou Bashir received his degree in Electronic Engineering in Algiers in 1978 and his
doctorate in 1990 at the University Claude Bernard Lyon 1 - France. Currently, he is
assistant professor at the University Badji Mokhtar, Annaba. His interests are in
intelligent systems.

IMPLEMENTATION OF THE DEVELOPMENT OF A FILTERING ALGORITHM TO IMPROVE THE SYSTEM OF HEARING IN HEARING IMPAIRED WITH COCHLEAR IMPLANT

More Related Content

What's hot (18)

Viewers also liked (20)

Similar to IMPLEMENTATION OF THE DEVELOPMENT OF A FILTERING ALGORITHM TO IMPROVE THE SYSTEM OF HEARING IN HEARING IMPAIRED WITH COCHLEAR IMPLANT (20)

Recently uploaded (20)

IMPLEMENTATION OF THE DEVELOPMENT OF A FILTERING ALGORITHM TO IMPROVE THE SYSTEM OF HEARING IN HEARING IMPAIRED WITH COCHLEAR IMPLANT