Study
and Implementation of Voiceprint Identity
Au
thentication for Android Mobile Terminal
Zh
en Huang
1,2
, Xiao
mei Zhang
1,
2
, Le
i Wang
1,3
, Zh
engying Li
1,
2
1
Schoo
l of Information Engineering, Wuhan University of Technology, Wuhan 430070, China
2
Key Laboratory of Fiber Optic Sensing Technology and Information Processing, Ministry of Education, Wuhan 430070, China
3
Hubei
Key Laboratory of Broadband Wireless Communication and Sensor Networks, Wuhan 430070, China
zhenhuang@whut.edu.cn,
may
125z@whut.edu.cn,
le
innawang@163.com,
zh
yli@whut.edu.cn
Abstract—With the popularity of the mobile terminals, identity
authentication for the mobile terminal applications has become a
key issue in protecting the security of the mobile terminal
information. The voiceprint recognition technology has broad
application prospects because of its flexible, economical and
practical advantages. This paper describes the realization of the
voiceprint identity authentication system for Android mobile
terminal in detail, including the design of the voiceprint
registration module and the design of the voiceprint recognition
module. In order to solve the problem that the training sample
data generalization ability is not high, this paper presents a
training and recognition method of voiceprint model combined
with stacking and cross validation. Finally, the performance of the
system is tested. The experimental results show that the system has
better authentication ability.
Keywords—Voiceprint Recognition; MFCC; SVM; Android;
Identity Authentication.
I. I
NTRODUCTION
With
the continuous development and progress of the
hacking technology, the traditional text password authentication
method due to the simple processing logic and realization
process , compared with the biometric based authentication
method, its security is clearly far less than the latter. Everyone's
vocal organs, sound habits, sound channels differ a lot, hence
each person's voiceprints is different. So the use of voice
characteristics of identity authentication as a viable option [1].
The use of voiceprint features for identity authentication in
addition to not forgetting, will not be lost, easy to use and other
biometric identity methods such as the common advantages, but
also easy to accept, low cost, long-range identity authentication
and other advantages [2], so that in the field of finance [3],
national defense and military fields [4], public security and
judicial fields [5] have a good application.
With the development of LTE, the mobile terminals have
been developing rapidly. According to the data released by the
US survey company IDC shows that in 2016 the global total
sales of smart phones reached 1.47 billion, of which the share of
Android is more than 87% and still show a gradual upward trend
[6]. Because Android-based mobile terminal comes with voice
acquisition equipment and it has the ability of dealing with data,
so the application of voiceprint recognition technology for
identity authentication in the mobile terminal has a good
prospects of application. In this paper, the Android mobile
terminal as the realization platform, select MFCC as the voice
feature parameters, choose the SVM training voiceprint model,
and the use of stacking integrated learning method of voice
feature parameters of the training and identification process
design, design and implementation of the voiceprint the
application of identity authentication system on Android
platform.
The rest of this paper is organized as follows: Section II
presents the design of the system, including the design of the
voiceprint registration module and the design of the voiceprint
recognition module. Section III shows the experiment results.
And we conclude this paper in section IV.
II. T
HE
D
ESI
GN OF
T
HE
S
YSTEM
A. Desig
n of Voiceprint Registration Module
The registration of voiceprint is the basis of the entire
process of voiceprint identification, and the process of
identification relies on the voiceprint model generated by the
process of registration. The voiceprint registration module is
mainly composed of two parts: the verification user name and
the training voiceprint model. Its main process including the
user name verification, the audio acquisition, the voiceprint
feature extraction and the voiceprint model training and other
processes .The flow chart of voiceprint registration module is
shown in Fig. 1.
Re
gistered user
name
Whether the user
name exists
Spec
ify and
return flags
Yes
Reminder user
name is
registered
Re
cord
voice
Add a flag for
the voiceprint
feature
Training SVM
feature model
Ex
traction of voice feature
parameters
Pretreatment Extract MFCC
Add
a
flag
No
Training
Voiceprint model
Fig. 1. Th
e flow chart of voiceprint registration module
2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics( CISP-BMEI 2017)
978-1-5386-1936-0/17/$31 ©2017 IEEE