Method and system for denoising high-frequency ultrasound based on multipath matching pursuit algorithm

Patent metadata

Field	Value
Publication number	US-12379353-B2
Application number	US-202418627342-A
Country	US
Kind code	B2
Filing date	Apr 4, 2024
Priority date	Jun 30, 2022
Publication date	Aug 5, 2025
Grant date	Aug 5, 2025

Abstract

The present invention provides an improved method and system for denoising high-frequency ultrasound based on a multipath matching pursuit algorithm. The method includes: acquiring a high-frequency ultrasound detection signal of a to-be-tested sample; constructing a discrete overcomplete dictionary according to the high-frequency ultrasound detection signal, and training the discrete overcomplete dictionary; reconstructing the high-frequency ultrasound detection signal by using a trained dictionary and using a multipath matching pursuit algorithm, and obtaining a global optimal atom; performing interpolation on the global optimal atom, and constructing a consecutive atomic library; and reconstructing the high-frequency ultrasound detection signal in the consecutive atomic library according to a parameter of the global optimal atom, to complete signal denoising. In the present invention, a signal-to-noise ratio and detection precision of a high-frequency ultrasound signal are improved, so that a reflected signal and a position of a microdefect inside a sample can be observed more effectively.

First claim

What is claimed is:

1. An improved method for denoising high-frequency ultrasound based on a multipath matching pursuit algorithm, comprising:
S1: acquiring a high-frequency ultrasound detection signal of a to-be-tested sample;
S2: constructing a discrete overcomplete dictionary according to the high-frequency ultrasound detection signal, and training the discrete overcomplete dictionary;
S3: reconstructing the high-frequency ultrasound detection signal by using a trained dictionary and using a multipath matching pursuit algorithm, and obtaining a global optimal atom;
S4: performing interpolation on the global optimal atom, and constructing a consecutive atomic library; and
S5: reconstructing the high-frequency ultrasound detection signal in the consecutive atomic library according to a parameter of the global optimal atom, to complete signal denoising,
wherein the reconstructing the high-frequency ultrasound detection signal by using a trained dictionary and using a multipath matching pursuit algorithm, and obtaining a global optimal atom in step S3 specifically comprises steps of:
S31: giving a dictionary D, a to-be-processed signal y, and a sparsity k, initializing a sparse coefficient α 0 =0 and a residual r 0 =y, and reconstructing an atom set Ω 0 =ϕ and an index set ω 0 =ϕ;
S32: for a t th iteration, calculating a residual r t−1 , taking inner products of all atoms in the dictionary matrix D, finding G corresponding atoms with the largest inner products and corresponding indices, only keeping n paths according to each iteration, setting a threshold according to an inner product of a current optimal atom and the residual, and eliminating a selected atom if an inner product of the atom and a signal is less than the threshold;
S33: updating an index set ω t and a corresponding reconstructed atom set Ω t :

ω
t

=

[

ω

t
-
1

,

λ
t

]

,

Ω
t

=

[

Ω

t
-
1

,

]

;

wherein λ t represents a t th iteration index, represents an atom for the t th iteration;
S34: calculating a sparse coefficient α t of the reconstructed atom set Ω t corresponding to the to-be-processed signal y;
S35: determining whether an iteration termination condition is met, and if the iteration termination condition is met, stopping the iteration and restoring a sparse coefficient α by using the index set ω t , or if the iteration termination condition is not met, letting t=t+1, until the iteration ends; and
S36: selecting a path with the smallest residual from all the paths for output, to obtain the global optimal atom.

2. The improved method for denoising high-frequency ultrasound based on a multipath matching pursuit algorithm according to claim 1 , wherein a method for acquiring a high-frequency ultrasound detection signal of a to-be-tested sample in step S1 comprises:
completely immersing the to-be-tested sample in deionized water, scanning the to-be-tested sample by using a high-frequency ultrasound probe, and saving a high-frequency ultrasound detection signal acquired during scanning, wherein a focal plane of the high-frequency ultrasound probe is disposed on a bottom surface of the to-be-tested sample.

3. The improved method for denoising high-frequency ultrasound based on a multipath matching pursuit algorithm according to claim 1 , wherein a method for constructing a discrete overcomplete dictionary according to the high-frequency ultrasound detection signal in step S2 comprises:
selecting iteration parameters according to the high-frequency ultrasound detection signal, and constructing the discrete overcomplete dictionary.

4. The improved method for denoising high-frequency ultrasound based on a multipath matching pursuit algorithm according to claim 3 , wherein the iteration parameters comprise a dictionary matrix D ∈ R M×X , a coefficient matrix αϵR k×M , an index set ω m , and a residual E m .

5. The improved method for denoising high-frequency ultrasound based on a multipath matching pursuit algorithm according to claim 1 , wherein the performing interpolation on the global optimal atom, and constructing a consecutive atomic library in step S4 specifically comprises steps of:
constructing the consecutive atomic library near a frequency of the global optimal atom by using polar coordinate interpolation, specifically represented as follows:
a starting atom

d
⁡
(

f
a

)

=

d
⁡
(

f
n

-

Δ
2

)

,

an initial atom d(f b )=d(f n ), and an ending atom

d
⁡
(

f
c

)

=

d
⁡
(

f
n

+

Δ
2

)

,

d
⁡
(

f
a

)

=

c
⁡
(

f
n

)

+

r
⁢

cos
⁡
(
θ

Assignees

• Univ Jiangnan
• China Key System & Integrated Circuit Co Ltd

Inventors

• Li Ke
• Ming Xuefei
• Su Lei
• Gu Jiefei
• Zhao Xinwei

Classifications

• G01N2291/0289

  Internal structure, e.g. defects, grain size, texture · CPC title

• G01N29/449

  Statistical methods not provided for in G01N29/4409, e.g. averaging, smoothing and interpolation · CPC title

• G01N29/4418

  with a model, e.g. best-fit, regression analysis · CPC title

• G01N29/28

  providing acoustic coupling {, e.g. water (impedance matching G10K11/02)} · CPC title

• G01N2291/044

  Internal reflections (echoes), e.g. on walls or defects · CPC title