Signal processing method in cochlear implant
A signal processing method in cochlear implant is performed by a speech processor and comprises a noise reduction stage and a signal compression stage. The noise reduction stage can efficiently reduce noise in a electrical speech signal of a normal speech. The signal compression stage can perform good signal compression to enhance signals to stimulate cochlear nerves of a hearing loss patient. The patient who uses a cochlear implant performing the signal processing method of the present invention can understand normal speech.
1. A signal processing method for a cochlear implant, wherein the cochlear implant comprises a microphone and a speech processor, the microphone is configures to convert sound into an electrical speech signal y, the speech processor is configured to receive the electrical speech signal y, process the electrical speech signal y, and output the processed electrical speech signals; the signal processing method comprising:
a noise reduction stage performed by the speech processor, comprising:
receiving the electrical speech signal y from the microphone;
segmenting the electrical speech signal y to a plurality of continuous noisy frames, the plurality of continuous noisy frames comprising a first noisy frame y 1 , a second noisy frame y 2 , . . . , a t-th noisy frame y t , . . . , and T-th noisy frame y T wherein T is a length of the sound;
reducing noises in the t-th noisy frame y t to obtain a t-th clean frame x t ; and
outputting the t-th clean frame x t ; and
a signal compression stage performed by the speech processor, comprising:
receiving an amplitude envelope of the t-th clean frame x t ;
compressing the amplitude envelope of the t-th clean frame x t to form a t-th output frame z t =α t ×(x t − x t )+ x t , wherein
x t is a mean of the amplitude envelope of the t-th clean frame x t ;
α t is a compression factor;
when the t-th output frame z t is in a monitoring range between an upper boundary and a lower boundary, α t =α t−1 +Δα 1 and Δα 1 is a positive value; and
when the t-th output frame z t is beyond the monitoring range, α t =α t−1 +Δα 2 and Δα 2 is a negative value; and
outputting the t-th output frame z t .
2. The signal processing method as claimed in claim 1 , wherein the t-th clean frame x t in the noise reduction stage is expressed as
x t =W 2 h ( y t )+ b 2
where
h(y t ) is a function including W 1 and b 1 in time domain;
W 1 and W 2 are default connection weights in the time domain; and
b 1 and b 2 are default vectors of biases of hidden layers of a DDAE (deep denoising autoencoder)-based NR (noise reduction) structure in the time domain.
3. The signal processing method as claimed in claim 2 , wherein h(y t ) in the noise reduction stage is expressed as
h
(
y
t
)
=
1
1
+
exp
[
-
(
W
1
y
t
+
b
1
)
]
.
4. The signal processing method as claimed in claim 3 , wherein the upper boundary in the signal compression stage is expressed as:
UB= x t +α 0 ×(max( x t )− x t ); and
the lower boundary in the signal compression stage is expressed as:
LB= x t +α 0 ×(min( x t )− x t ).
5. The signal processing method as claimed in claim 2 , wherein the upper boundary in the signal compression stage is expressed as:
UB= x t +α 0 ×(max( x t )− x t ); and
the lower boundary in the signal compression stage is expressed as:
LB= x t +α 0 ×(min( x t )− x t ).
6. The signal processing method as claimed in claim 1 , wherein the t-th clean frame x t , in the noise reduction stage is expressed as
x t =InvF {( W 2 ′h ′( F{y t })+ b 2 ′)}
where F { } is a Fourier transform function to transfer the t-th noisy frame y t from time domain to frequency domain;
h′ ( ) is a function including W 1 ′ and b 1 ′;
W 1 ′ and W 2 ′ are default connection weights in frequency domain;
b 1 ′ and b 2 ′ are default vectors of biases of hidden layers of a DDAE-based NR structure in frequency domain; and
InvF { } is an inverse Fourier transform function to obtain the t-th clean frame x t .
7. The signal processing method as claimed in claim 6 , wherein h′ (F {yt}) in the noise reduction stage is expressed as
h
′
(
F
{
y
t
}
)
=
1
1
+
exp
[
-
(
W
1
′
F
{
y
t
}
+
b
1
′
)
]
.
8. The signal processing method as claimed in claim 7 , wherein the upper boundary in the signal compression stage is expressed as:
UB= x t +α 0 ×(max( x t )− x t ); and
the lower boundary in the signal compression stage is expressed as:
LB= x t +α 0 ×(min( x t )− x t ).
9. The signal processing method as claimed in claim 6 , wherein the upper boundary in the signal compression stage is expressed as:
UB= x t +α 0 ×(max( x t )− x t ); and
the lower boundary in the signal compression stage is expressed as:
LB= x t +α 0 ×(min( x t )− x t ).
10. The signal processing method as claimed in claim 1 , wherein the upper boundary in the signal compression stage is expressed as:
UB= x t +α 0 ×(max( x t )− x t ); and
the lower boundary in the signal compression stage is expressed as:
LB= x t +α 0 ×(min( x t )− x t ).