WO2020042219A1 - 一种用于可穿戴设备的新型身份验证方法 - Google Patents

一种用于可穿戴设备的新型身份验证方法 Download PDF

Info

Publication number
WO2020042219A1
WO2020042219A1 PCT/CN2018/105076 CN2018105076W WO2020042219A1 WO 2020042219 A1 WO2020042219 A1 WO 2020042219A1 CN 2018105076 W CN2018105076 W CN 2018105076W WO 2020042219 A1 WO2020042219 A1 WO 2020042219A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
data
vibration
processing
length
Prior art date
Application number
PCT/CN2018/105076
Other languages
English (en)
French (fr)
Inventor
伍楷舜
陈文强
王璐
杨宜坚
Original Assignee
深圳大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳大学 filed Critical 深圳大学
Publication of WO2020042219A1 publication Critical patent/WO2020042219A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
    • G06F21/31—User authentication
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
    • G06F21/45—Structures or tools for the administration of authentication

Definitions

  • the present invention relates to the field of information processing technology, and in particular, to a novel identity verification method for a mobile device.
  • wearable smart sensing devices are rapidly developing.
  • hand-held devices such as smart bracelets and smart watches are also quite popular.
  • the identification of wearable devices is very dependent on mobile phones and computers.
  • the purpose of wearable device identification is to distinguish legitimate users from illegal users, thereby protecting the relevant rights and interests of legitimate users, such as protecting property security and protecting private information.
  • identity recognition is even more important.
  • the present invention provides a new identity verification method for wearable devices. Based on bone conduction vibration and machine learning, the existing accelerometers and gyroscopes of wearable devices are used to provide a Intelligent identification method, low hardware cost, simple equipment and system, easy to use, suitable for most wearable devices on the market.
  • a new authentication method for a wearable device includes the following steps:
  • the wearable device When the user taps on the back of the hand, the wearable device records the three-axis acceleration data and three-axis angular velocity data detected by the accelerometer and gyroscope built in the wearable device;
  • the user taps the new unlock signal data as test data, checks whether the test data is a valid tap, and then uses the method of processing training data to process the test data;
  • the processed test data can be classified with the data of the training set by a machine learning classification method, so as to perform identification verification.
  • step S1 the user of the wearable device taps the hand wearing the wearable device; the tapping action means that the finger of the other hand hits the back of the hand briefly and quickly. exercise.
  • step S1 the user of the smart watch taps the back of the hand multiple times; there are time intervals between two consecutive taps.
  • the time interval between two adjacent times can make the values recorded by the accelerometer and gyroscope more accurate.
  • the step S2 includes: S21: performing filtering processing on the data detected by the accelerometer and the gyroscope to obtain acceleration data and angular velocity data.
  • S22. Perform slice processing on the filtered acceleration data and angular velocity data, and only take out the peaks of equal length and the acceleration data and angular velocity data in the vicinity, and remove the data without vibration signals.
  • the step S3 includes: S31. Perform alignment processing on the vibration signals after the end-segment segmentation through the overall cross-correlation method.
  • the specific operation of the alignment processing is to calculate the offset between the two vibration signals. Measure, and then move the current vibration signal, after the movement, only the complete part shared between the two vibration signals is taken;
  • O (A, B) P (A, B) -n calculates the offset O (A, B) between two vibration signals, where a and b represent two vibration signals with a length of n, a (i) represents the amplitude of the i-th point of the vibration signal a, b (i) represents the amplitude of the i-th point of the vibration signal b, and C (a, b) represents the correlation between the vibration signal a and the vibration signal b Degree; A means zero-padded the length n on both sides of the vibration signal a to obtain a first signal of length 3n; B means vibration signal b of length n; P (A, B) means the first signal A The signal position of length n having the highest correlation with the second signal B is the position of n; O (A, B) is the offset between the calculated first signal A and the second signal B.
  • step S4 a power spectral density feature of the vibration signal after the alignment process is extracted, and the power spectral density feature and the amplitude feature of the vibration signal before the alignment process are collectively used as the extracted signal feature.
  • the step S5 includes: S51.
  • the user taps new unlock signal data as test data, and checks whether the test data is a valid tap.
  • the valid tap signal should first be a gentle signal. This is followed by a signal peak, followed by a gentle signal. Check whether the value of several signal points before and after the signal exceeds the set threshold, if it exceeds the threshold, discard the collected signal; S52.
  • Process the test data by processing the training data, that is, systematically filter the test data Noise reduction and endpoint segmentation processing, then alignment processing with the training data, and finally fast Fourier transform of the segmented processing training data.
  • the step S6 includes: S61.
  • Multi-key combination authentication The tapping signal data of the user tapping the articulating bones of four fingers except the thumb is used as training data, and it is stored in the database. And the location of the multi-key stroke is also saved as a combination password.
  • S62. When the user performs identity verification, each tapping signal is compared and classified with all the original tapping signals in the database through a machine learning algorithm KNN, so as to determine the back key position corresponding to each tapping by the user. Multiple keystroke combination passwords are compared with the original database passwords for verification.
  • One-click tapping authentication The user only needs to tap a certain position on the back of the hand repeatedly to store it as training data in the database.
  • the generated tap signal is compared with the training data stored in the database. Calculate the Euclidean distance between the test data and the training data. If the obtained distance does not exceed the set threshold, it is determined as a legal tap for verification.
  • the principle of the above two identification methods is that the vibration signals generated by tapping on different human bodies or different positions of the same human body are different, and using different vibration signals can be used to distinguish the tapping of different users, thereby Implement authentication.
  • step S2 a Butterworth filter is used to perform filtering and noise reduction processing on the collected vibration signal, and a high-pass filter with a cut-off frequency of 20 Hz is used to filter out DC components and low-frequency noise. 300Hz low-pass filtering removes high-frequency noise.
  • step S2 the endpoint segmentation process uses a fixed length to traverse a whole section of vibration signal.
  • the energy of this segment of signal is the largest, it is considered that the tap signal appears, take The length of the segment and a signal of a certain length before and after it are used as the vibration signal after the end segment is cut.
  • the beneficial effect of the present invention is that compared with the prior art, the present invention provides a new wearable device identity verification method.
  • the vibration signal generated by hitting the back of the hand is converted into a digital signal and stored in the watch.
  • the beneficial effects of the invention are low hardware cost, simple equipment and system, and convenient use. Because wearable devices have a wide range of applicable populations, they have significant improvements compared to existing technologies.
  • FIG. 1 is a schematic flowchart of a novel authentication method for a wearable device according to the present invention
  • FIG. 2 is an equivalent diagram of vibration transmitted on the back of the hand when the back of the hand is hit;
  • FIG. 3 is a schematic diagram of an effect simulation before the alignment process is implemented according to the present invention.
  • FIG. 4 is a schematic diagram of an effect simulation after the alignment process is implemented according to the present invention.
  • FIG. 5 is a schematic view showing an effect of a virtual key position of the present invention.
  • the wearable device's identity verification method based on bone conduction vibration and machine learning includes the following steps:
  • the wearable device When the user taps on the back of the hand, the wearable device records the three-axis acceleration data and three-axis angular velocity data detected by the accelerometer and gyroscope built in the wearable device;
  • the user taps the new unlock signal data as test data, checks whether the test data is a valid tap, and then processes the test data by processing the training data;
  • the processed test data can be classified with the data of the training set by a machine learning classification method, so as to perform identification verification.
  • this embodiment implements input on the back of the hand by using the principle of bone conduction vibration, that is, the back of the hand is used as an input device (the virtual key can be any position on the back of the hand) to implement the user's input function, and
  • the vibration signal generated by the hitter's back lock is the unlock code. It is very convenient to use vibration to realize identity recognition.
  • step S1 the user of the wearable device taps the hand wearing the wearable device; the tapping action refers to a short and rapid movement of the finger of the other hand on the back of the hand.
  • the wearable device detects a vibration signal of the user tapping the back of the hand and converts the signal into a digital signal for processing. There are time intervals between two adjacent strikes. Having a time interval between two adjacent times can make the values recorded by the accelerometer and gyroscope more accurate.
  • the raw data collected by the accelerator and the gyroscope includes noise interference caused by human movement and other reasons. Therefore, first, a denoising process is required to make the signal more effective. Unlike the signal characteristics of radio frequency signals or sound signals that need to travel through space, vibration signals are less affected by surrounding environmental noise. Therefore, a reasonable filter can be used to remove low frequency and high frequency noise outside the frequency of the vibration signal. Desired purpose.
  • the frequency of the vibration signal generated by striking the human body is generally 20 to 300 Hz.
  • a Butterworth high-pass filter is used to filter the DC component in the signal and low-frequency noise generated by the movement of the human body itself (the frequency of the noise is usually lower than 5HZ), and then a Butterworth low-pass filter is used to filter In addition to high-frequency components.
  • the filtered signal will be used to extract a valid tap signal.
  • the endpoint segmentation processing is also referred to as endpoint detection processing.
  • the processing process uses a fixed length to traverse the entire section of the vibration signal. When the energy of this section of the signal is the largest, it is considered that the tap signal appears. , Take the signal of the length of the segment and a certain length before and after it as the vibration signal after the end segment is cut.
  • the vibration signal after the end segment is also called the knock signal.
  • a general cross-correlation method (general cross-correlation (GCC)) is used to perform alignment processing on the vibration signals after segmentation at the endpoints.
  • GCC general cross-correlation
  • the specific operation of the alignment processing is to calculate the offset between the two vibration signals. Shift, and then move the current vibration signal. After the movement, only the complete part shared between the two vibration signals is taken.
  • the alignment processing described in this embodiment can align all vibration signals, which is beneficial to the improvement of the classification accuracy of the machine learning algorithm.
  • the simulation results before and after the alignment processing are shown in FIG. 3 and FIG. 4.
  • step S3 described in this embodiment except for the first strike, the other strike signals are aligned with the first strike signal, thereby ensuring that the vibration data in the corresponding dimensions of all the strike signals are one by one. correspond.
  • step S3 described in this embodiment the formula
  • O (A, B) P (A, B) -n calculates the offset O (A, B) between two vibration signals, where a and b represent two vibration signals with a length of n, a (i) represents the amplitude of the i-th point of the vibration signal a, b (i) represents the amplitude of the i-th point of the vibration signal b, and C (a, b) represents the correlation between the vibration signal a and the vibration signal b Degree; A means zero-padded the length n on both sides of the vibration signal a to obtain a first signal of length 3n; B means vibration signal b of length n; P (A, B) means the first signal A The signal position of length n having the highest correlation with the second signal B is the position of n; O (A, B) is the offset between the calculated first signal A and the second signal B.
  • a power spectral density (PSD) feature of the vibration signal after the alignment process is extracted, and the power spectral density feature and the amplitude feature of the vibration signal before the alignment process are collectively used as the extracted Signal characteristics.
  • PSD power spectral density
  • a formula is used Extract the power spectrum density characteristic PSD of the vibration signal after alignment processing, where f s is the sampling frequency of the vibration signal, n is the signal length, k is the signal with signal length n, and FFT (k) is the Fourier of signal k Leaf transformation, abs (FFT (k)) means to take the absolute value of FFT (k).
  • the power spectral density describes the signal power characteristics of the vibration signal at a series of frequency points.
  • the weighted PSD is called a weighted PSD.
  • both position-sensitive frequency points and position-related frequency points can provide valuable references.
  • the fisher score technology can be used to classify these two frequency points into one class for identification, and the calculation formula is:
  • r represents the frequency position
  • n i represents the number of samples of the i-th class
  • u i represents the mean and variance of the samples of the i-th class
  • u represents the mean of all classes of the feature of this dimension.
  • the system only needs to identify whether it is a vibration generated by tapping at a fixed position, and the position-related frequency points will exhibit better characteristics.
  • E ( ⁇ ) is the variance.
  • the frequency points with small variance represent positional correlation and can get greater weight; on the other hand, the dimensional frequency points with large variance are considered as poor description features and are multiplied by smaller weights.
  • step S4 of this embodiment a predetermined number of training samples are collected at each back of the hand position, corresponding signal features are extracted, and the signal features and labels of the training samples are stored in the database as a training set.
  • the predetermined number can be customized and adjusted according to the needs of the user, and the predetermined number of each position in this embodiment is preferably 30.
  • step S5 in this embodiment the user taps the new unlock signal data as test data, and checks whether each test data is a valid tap, because unconscious tap vibrations such as arm swings may trigger.
  • the execution of the authentication algorithm and the filtering of invalid tap signals can reduce the program power consumption. If the vibration signal is a swing of the arm, the vibration signal should be a continuous vibration wave. After filtering, the effective tapping signal should first be a smooth signal, then the signal peak, and finally a smooth signal. It is detected whether the value of several signal points before and after the signal exceeds a set threshold. If the value exceeds the threshold, the collected signal is discarded.
  • step S5 in this embodiment the method of processing training data is used to process the test data, that is, segmentation, endpoint alignment processing, and then alignment processing with the training data, and finally fast Fourier transform and calculation of power spectral density calculation.
  • step S6 described in this embodiment there are two methods for performing identity verification: multi-key identification and single-key identification.
  • multi-key recognition the vibration signal that the user taps for unlocking is compared with the training data stored in the original 4 locations, and the k-NearestNeighbor algorithm is used to classify the location of the tap. Multiple tap locations are combined to compare with the unlock password stored in the database to enable identity verification.
  • step S6 described in this embodiment in one-button recognition, the vibration signal generated by the user's tapping is compared with the original training data, and the Euclidean distance from each training data is calculated. Threshold, the test data is considered to be similar to the training data. (The threshold is dynamically obtained by calculating the distance between two pairs of training data.) If the training data is similar to the test data, it can be considered that the tap is legal, so as to perform identity recognition.
  • Equation (1) can be further interpreted as
  • the vibration during finger tapping can be divided into two phases.
  • the first stage there is an instantaneous contact ⁇ t between the finger and the human body, which is considered to be a forced vibration of a constant force F (0).
  • F (0) the contact between the finger and the skin surface disappears, causing the system to vibrate on its own, which is called free vibration.
  • X (w) is the frequency spectrum of vertical vibration
  • w is the frequency. It can be seen from the formula (4) that X (w) is jointly determined by the striking force F (0), mass m, damping coefficient c, and spring constant k.
  • the initial vibration signal x (t) based on the human body's striking is strongly related to the strength of the striking body and the structure of the human body. Therefore, the difference between the striking person and the striking position will make x (t The signal characteristics are different.
  • the vibration signal x (t) will be attenuated during the propagation from the finger tap position to the sensor.
  • the model can be expressed as:
  • y (t) is the vertical displacement at the location where the vibration propagates
  • d is the propagation distance
  • ⁇ is the attenuation coefficient related to the propagation medium and frequency
  • ⁇ f ( ⁇ , w)
  • ⁇ represents the density of the propagation medium.
  • the vibration signal When a certain part of the human body is struck, due to the complex physiological structure of the human body, the vibration signal will be transmitted through substances of different densities in the human body, forming a composite signal composed of multiple paths at the location of the sensor, namely:
  • vibration signals have specific attenuation characteristics when they are transmitted at different positions, and each tapping action will also exhibit specific signal characteristics at various positions on the human body. It is the inconsistency of the tap signals at different locations that can implement the authentication function.
  • the present invention provides a new wearable device authentication method.
  • the vibration signal generated by hitting the back of the hand is converted into a digital signal and stored in the watch.
  • the invention can be used for unlocking smart wearable devices, mobile payments, and the like.
  • the beneficial effects of the invention are low hardware cost, simple equipment and system, and convenient use. Because wearable devices have a wide range of applicable populations, they have significant improvements compared to the prior art.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Security & Cryptography (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Software Systems (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

一种用于可穿戴设备的身份验证方法，包括：在用户敲击手背时，记录内置于可穿戴设备中加速度计和陀螺仪所检测到的三轴加速度数据和三轴角速度数据；对采集到的信号进行处理得到频域上的数据，拼接在原来的数据上；将处理完的信号数据作为训练数据集储存在可穿戴设备数据库里；用户敲击新的解锁信号数据作为测试数据，先检验测试数据是否为一次有效的敲击，然后用处理训练数据的方法来处理测试数据；处理完的测试数据可以与训练集的数据通过机器学习分类方法进行分类，从而进行身份验证的识别。其硬件成本低，用户只需要敲击手背即可解锁可穿戴设备，学习成本低，使用方便，交互方式新颖有趣且方便快捷，应用广泛。

Description

一种用于可穿戴设备的新型身份验证方法 技术领域
本发明涉及信息处理技术领域，尤其涉及一种移动设备的新型身份验证方式。
背景技术
目前，可穿戴智能感知设备迅速发展，其中智能手环和智能手表等手戴式设备也颇为流行，但由于其大小，成本等限制，可穿戴设备的身份识别却很依赖于手机，电脑。可穿戴设备的身份识别的目的在于区分合法用户与非法用户、从而保护合法用户的相关权益，如保障财产安全、保护隐私信息。在如今移动支付技术盛行、信息交换便利的时代背景下，身份识别具有更加重要的意义。
传统智能终端的身份识别方式有文本密码识别、网格锁识别等，均要求用户记住长串密码或网格形状，且存在容易泄密等安全问题，在可穿戴设备的小屏幕上很容易误触，用户体验不佳。语音认证利用用户声音特征进行识别，这容易受到环境噪声的影响。而基于生物信息图像识别的身份认证方案如指纹识别、人脸识别、虹膜识别，因为可穿戴设备的体积过小，成本等原因受到限制。
发明内容
为了解决现有技术中的问题，本发明提供了一种用于可穿戴设备的新型身份验证方法，基于骨传导振动与机器学习，利用可穿戴设备现有的加速度计和陀螺仪，提供一种智能的身份识别方法，硬件成本低，设备和系统简单，使用方便，适用于目前市面上绝大部分的可穿戴设备。
本发明具体通过如下技术方案实现：
一种用于可穿戴设备的新型身份验证方法，包括以下步骤：
S1、可穿戴设备在用户敲击手背时，记录内置于可穿戴设备中加速度计和陀螺仪所检测到的三轴加速度数据和三轴角速度数据；
S2、对采集到的加速度信号和角速度信号进行滤波降噪和端点切段处理；
S3、对端点切断后的加速度信号和角速度信号进行对齐处理；
S4、将对齐后的加速度数据和角速度数据进行快速傅里叶变换得到频域上的数据，拼接在原来的数据上；将处理完的信号数据作为训练数据集储存在可穿戴设备的数据库里；
S5、用户敲击新的解锁信号数据作为测试数据，检验测试数据是否为一次有效的敲击，然后用处理训练数据的方法来处理测试数据；
S6、处理完的测试数据可以与训练集的数据通过机器学习分类方法进行分类，从而进行身份验证的识别。
作为本发明的进一步改进，在步骤S1中，可穿戴设备的使用者敲击的是佩戴可穿戴设备的那只手；所述的敲击动作是指另外一只手手指敲击手背短暂而快速的运动。
作为本发明的进一步改进，在步骤S1中，所述的智能表使用者多次敲击手背；相邻两次敲击留有时间间隔。相邻两次留有时间间隔可以使加速度计和陀螺仪记录的值更准确。
作为本发明的进一步改进，所述步骤S2包括：S21、分别对所述加速度计和陀螺仪检测到的数据进行滤波处理，获得加速度数据和角速度数据。S22、将获得滤波后的加速度数据和角速度数据进行切片处理，只取出等长的峰值及其附近的加速度数据和角速度数据，去掉没有振动信号的数据。
作为本发明的进一步改进，所述步骤S3包括：S31、通过总体互相关法对端点切段后的振动信号进行对齐处理，所述对齐处理的具体操作是计算两个振动信号之间的偏移量，然后对当前的振动信号进行移动，移动完之后只取两个振动信号之间共有的完整部分；
S32、对齐处理通过公式
Figure PCTCN2018105076-appb-000001
以及O(A,B)＝P(A,B)-n计算两个振动信号之间的偏移量O(A,B)，其中，a和b代表两个信号长度为n的振动信号，a(i)表示振动信号a的第i个点的振幅大小，b(i)表示振动信号b的第i个点的振幅大小，C(a,b)表示振动信号a和振动信号b的相关度；A表示对振动信号a两边长度为n的部分进行补零，进而获得的一个长度为3n的第一信号；B表示长度n的振动信号b；P(A,B)表示第一信号A中与第二信号B相关度最高的长度为n的信号位置；O(A,B)为计算所得的第一信号A与第二信号B之间的偏移量。
作为本发明的进一步改进，所述步骤S4中，提取对齐处理后的振动信号的功率谱密度特征，并将所述功率谱密度特征与对齐处理前振动信号的振幅特征共同作为提取的信号特征。
作为本发明的进一步改进，所述步骤S5包括：S51、用户敲击新的解 锁信号数据作为测试数据，检验测试数据是否为一次有效的敲击，有效的敲击信号应该先是一段平缓的信号，接着是信号峰，最后是一段平缓的信号。检测信号前后若干个信号点的值是否超过设定的阈值，如果超过阈值，则丢弃该次采集的信号；S52、用处理训练数据的方法来处理测试数据，也就是对测试数据进行系统的滤波降噪和端点切段处理，然后与训练数据进行对齐处理，最后对切段处理的训练数据进行快速傅里叶变换。
作为本发明的进一步改进，所述步骤S6包括：S61、多键组合身份验证：用户敲击除了大拇指外的4手指的关节骨的敲击信号数据作为训练数据，将其保存在数据库里。并且将多键敲击的位置作为组合密码也一并保存起来。S62、用户进行身份验证时的，每一次的敲击信号与数据库原有的所有敲击信号通过机器学习算法KNN进行对比分类，从而确定用户每一次敲击所对应的手背键位。多个键敲击组合密码与数据库原密码对比，从而进行验证。S63、单键敲击身份验证：用户只需重复敲击手背某一个位置作为训练数据保存在数据库里。S64、用户进行身份验证时，只需敲击手背一下，所产生的敲击信号与数据库保存的训练数据进行比较。计算测试数据与训练数据的欧氏距离，若所得距离不超过所设定的阈值则判定为合法敲击，从而进行验证。
上述两种身份识别方法的原理在于在不同的人体，或者相同人体不同位置上敲击所产生的振动信号是不同的，而利用不同的振动信号，可以用于区分不同使用者的敲击，从而实现身份验证。
作为本发明的进一步改进，所述步骤S2中：采用巴特沃兹滤波器对采集的振动信号进行滤波降噪处理，使用截止频率为20Hz的高通滤波滤除直流分量和低频噪音，使用截止频率为300Hz的低通滤波滤除高频噪音。
作为本发明的进一步改进，所述步骤S2中：S2、所述端点切段处理中，使用固定长度来遍历整一段振动信号，当该段信号的能量最大时，则认为敲击信号出现，取该段长度及其前后一定长度的信号作为端点切段后的振动信号。
本发明的有益效果是：相比于现有技术，本发明提供一种新型的可穿戴设备身份验证方式。通过利用智能手机内置加速度计和陀螺仪，将敲击手背产生的振动信号转化为数字信号，保存在手表中。在进行身份验证时利用机器学习算法进行分类判断。本发明的有益效果在于硬件成本低，设备和系统简单，使用方便。由于可穿戴设备有广泛的适用人群，相比于现 有技术具有重要的改进意义。
附图说明
图1是本发明的可穿戴设备的新型身份验证方法流程示意图；
图2是敲击手背时振动在手背上传递的等效图；
图3是本发明实现对齐处理前的效果仿真示意图；
图4是本发明实现对齐处理后的效果仿真示意图；
图5是本发明的虚拟键位的效果示意图。
具体实施方式
下面结合附图说明及具体实施方式对本发明进一步说明。
本发明的可穿戴设备的基于骨传导振动与机器学习的身份验证方法，如图1所示，包括以下步骤：
S1、可穿戴设备在用户敲击手背时，记录内置于可穿戴设备中加速度计和陀螺仪所检测到的三轴加速度数据和三轴角速度数据；
S2、对采集到的加速度信号和角速度信号进行滤波降噪和端点切断处理；
S3、对端点切断后的加速度信号和角速度信号进行对齐处理；
S4、将对齐后的加速度数据和角速度数据进行快速傅里叶变换得到频域上的数据，拼接在原来的数据上；将处理完的信号数据作为训练数据集储存在可穿戴设备的数据库里；
S5、用户敲击新的解锁信号数据作为测试数据，检验测试数据是否为一次有效的敲击，然后以处理训练数据的方法处理测试数据；
S6、处理完的测试数据可以与训练集的数据通过机器学习分类方法进行分类，从而进行身份验证的识别。
如图5所示，本实施例通过骨传导振动原理实现在手背上进行输入，即把手背当作一个输入设备(该虚拟按键可以是手背的任意一个位置)来实现用户的输入功能，而敲击手背锁产生的振动信号就是解锁密码。利用振动来实现身份识别是非常便捷的。
具体地，在步骤S1中，可穿戴设备的使用者敲击的是佩戴可穿戴设备的那只手；所述的敲击动作是指另外一只手手指敲击手背短暂而快速的运动。可穿戴设备检测用户敲击手背的振动信号并将该信号转化为可供处理的数字信号。相邻两次敲击留有时间间隔。相邻两次留有时间间隔可以使加速度计和陀螺仪记录的值更准确.
本实施例所述步骤S2中，通过加速器和陀螺仪采集到的原始数据包含了人体移动等原因产生的噪声干扰，因此首先需要进行去噪处理，以使得信号更为有效。与射频信号或声音信号需要经过空间传播的信号特性不同，振动信号受到周围环境噪声的影响很小，因此，选用合理的滤波器去除振动信号所在频率之外的低频和高频噪声，即可达到所需目的。敲击人体产生的振动信号频率一般为20～300Hz。因此，本实施例采用巴特沃斯高通滤波器来滤除信号中的直流分量和人体本身移动产生的低频噪声(该噪声的频率通常低于5HZ)，然后采用巴特沃斯低通滤波器来滤除高频成分。经过滤波之后的信号将会用于提取有效的敲击信号。
本实施例所述步骤S2中，所述端点切段处理也称为端点检测处理，其处理过程使用固定长度来遍历整一段振动信号，当该段信号的能量最大时，则认为敲击信号出现，取该段长度及其前后一定长度的信号作为端点切段后的振动信号。端点切段后的振动信号也称为敲击信号。
本实施例所述步骤S3中，通过总体互相关法(general cross correlation，GCC)对端点切段后的振动信号进行对齐处理，所述对齐处理的具体操作是计算两个振动信号之间的偏移量，然后对当前的振动信号进行移动，移动完之后只取两个振动信号之间共有的完整部分。本实施例所述对齐处理处理能够将所有振动信号对齐，有利于机器学习算法分类精度的提升，其对齐处理前和对齐处理后的仿真效果图如图3和图4所示。
本实施例所述的步骤S3中，除第一次敲击外，其它敲击信号都是与第一次敲击信号进行对齐，从而保证所有敲击信号相应维度上的振动数据都是一一对应。
本实施例所述步骤S3中，通过公式
Figure PCTCN2018105076-appb-000002
以及O(A,B)＝P(A,B)-n计算两个振动信号之间的偏移量O(A,B)，其中，a和b代表两个信号长度为n的振动信号，a(i)表示振动信号a的第i个点的振幅大小，b(i)表示振动信号b的第i个点的振幅大小，C(a,b)表示振动信号a和振动信号b的相关度；A表示对振动信号a两边长度为n的部分进行补零，进而获得的一个长度为3n的第一信号；B表示长度n的振动信号b；P(A,B)表示第一信号A中与第二信号B相关度最高的长度为n的信号位置；O(A,B)为计算所得的第一信号A与第二信号B之间的偏移量。
本实施例所述步骤S4中，提取对齐处理后的振动信号的功率谱密度特征(power spectral density,PSD)，并将所述功率谱密度特征与对齐处理前振动信号的振幅特征共同作为提取的信号特征。优选的，所述步骤S4中，通过公式
Figure PCTCN2018105076-appb-000003
提取对齐处理后的振动信号的功率谱密度特征PSD，其中，f s为振动信号的采样频率，n为信号长度，k表示信号长度为n的信号，FFT(k)表示对信号k的傅里叶变换，abs(FFT(k))表示对FFT(k)取绝对值。
本实施例所述步骤S4中，功率谱密度描绘的是振动信号在一系列频点上的信号功率特征，通过对位置敏感频点或位置相关频点进行区别于其他频点的加权操作，可放大振动信号特征，提高特征识别准备度。加权操作后的PSD称为加权PSD。在键盘输入等需要获取多个振动位置信息的交互场景中，位置敏感频点及位置相关频点均可以提供有价值的参考。在本实施例中，利用fisher score技术可将这两种频点归为一类进行识别，其计算公式为：
Figure PCTCN2018105076-appb-000004
其中，r表示频点位置，n i表示第i类的样本个数，u i和
Figure PCTCN2018105076-appb-000005
代表第i类的样本的平均值和方差，u表示该维度特征的所有类的平均值。通过该技术计算出来的分数，我们设置成每个频点的权重，即w·X i，其中w＝F r，X是频谱，i是频谱里每个频点的序号。由此，不同位置振动信号每个频点的特征根据fisher score的分值得到了缩放，使得细粒度的可区分不同位置敲击振动的特征得到有效放大。
在本实施例身份验证等交互场景中，系统只需要识别出是不是某个固定位置敲击产生的振动，位置相关频点将会展现更好的特性。我们可以为每个频点设置一个权重，
Figure PCTCN2018105076-appb-000006
其中E(·)是方差。通过该权重设计，方差小的频点代表位置相关、可以获得更大的权重；反之，方差大的维度频点被认为是不好的描述特征而被乘以较小的权重。
本实施例所述步骤S4中，每个手背位置采集预定数量的训练样本，提取对应的信号特征，将训练样本的信号特征及其标签作为训练集保存在数据库里。所述预定数量可以根据用户的需求进行自定义设置和调整，本 实施例所述每个位置预定数量优选为30。
本实施例所诉步骤S5中，用户敲击新的解锁信号数据作为测试数据，检验每一次的测试数据是否为一次有效的敲击，因为手臂的摆动等非有意识的敲击振动都有可能触发身份验证算法的执行，对无效的敲击信号过滤能降低程序功耗。如果振动信号为手臂的摆动，振动信号应该为持续的振动波。在滤波后，有效的敲击信号应该先是一段平缓的信号，接着是信号峰，最后是一段平缓的信号。检测信号前后若干个信号点的值是否超过设定的阈值，如果超过阈值，则丢弃该次采集的信号。
本实施例所诉步骤S5中，用处理训练数据的方法来处理测试数据，也就是切段，端点对齐处理，然后与训练数据进行对齐处理，最后快速傅里叶变换和计算功率谱密度计算。
本实施例所述步骤S6中，进行身份验证的方法有两种：多键识别和单键识别。在多键识别中，用户敲击用于解锁的振动信号与原有的4个位置储存的训练数据比较，使用k-NearestNeighbor算法进行分类，从而确定敲击的位置。多个敲击位置组合在一起与数据库储存的解锁密码比较，从而实现身份验证。
在本实施例所述的步骤S6中，在单键识别中，用户敲击一下产生的振动信号与原有的训练数据相比较，计算与每一个训练数据的欧氏距离，如果小于设定的阈值，则认为测试数据与训练数据是相似的。(阈值通过计算训练数据两两之间的距离动态获得)如果训练数据与测试数据相似，可以认为该次敲击是合法的，从而进行身份识别。
上述两种身份识别方法的原理在于在不同的人体，或者相同人体不同位置上敲击所产生的振动信号是不同的。人体本身非常复杂，为了便于分析基本的敲击振动模型，首先构建一个图2所示的单自由度模型。在此模型中，质量单元由一个恒定的m表示，弹性单元被定义为k，而阻尼系数为c。根据牛顿第二运动定律，当外力施加到人体并发生垂直位移时，我们就有了
F(t)＝ma(t)+kx(t)+cv(t)
其中F(t)是外力，v(t)是速度，x(t)是垂直位移，c是阻尼系数，k是弹簧常数，m是质量。等式(1)可以进一步解释为
Figure PCTCN2018105076-appb-000007
事实上，手指敲击过程中的振动可分为两个阶段。在第一阶段，手指与人体之间存在瞬时接触Δt，这被认为是恒定力F(0)的强迫振动。在最初的瞬变干扰之后，在第二阶段，手指和皮肤表面的接触消失，使系统自行振动，这被称为自由振动。
在强制振动中，对方程的两侧进行傅里叶变换，得到：
Figure PCTCN2018105076-appb-000008
即：
Figure PCTCN2018105076-appb-000009
其中X(w)是垂直振动的频谱，w是频率。由公式(4)此可见，X(w)由敲击力F(0)、质量m、阻尼系数c以及弹簧常数k共同决定。也就是说，基于人体的敲击产生的初始振动信号x(t)与敲击力度、敲击位置的人体结构强相关，因此，敲击的人、敲击位置的不同均会使得x(t)信号特征不同。
在从手指敲击位置到传感器的传播过程中，振动信号x(t)将遭受衰减。在单一介质情况下该模型可以表述为：
y(t)＝x(t)e -αd
其中y(t)是振动传播到的位置处的垂直位移，d为传播距离，α是与传播介质及频率相关的衰减系数，α＝f(ρ,w)，其中ρ代表传播介质密度。对于人体来说，骨头密度较大，衰减系数α较小，而软组织密度较小，衰减系数α较大。
当敲击人体某个部位时，由于人体复杂的生理构造，振动信号会经由人体内不同密度的物质进行传播，在传感器所在位置形成一个由多条路径组成的合成信号，即：
Figure PCTCN2018105076-appb-000010
由于人体每个部位的内部结构均有其特殊性，因此振动信号在不同位置传播时都具备特定的衰减特性，每一次敲击动作在人体各个位置也会展现出特定的信号特征。正是不同位置的敲击信号不一致性，从而可以实现身份验证功能。
相比于现有技术，本发明提供一种新型的可穿戴设备身份验证方式。通过利用智能手机内置加速度计和陀螺仪，将敲击手背产生的振动信号转 化为数字信号，保存在手表中。在进行身份验证时利用机器学习算法进行分类判断。本发明可以用于解锁智能穿戴设备，移动支付等。本发明的有益效果在于硬件成本低，设备和系统简单，使用方便。由于可穿戴设备有广泛的适用人群，相比于现有技术具有重要的改进意义。
以上内容是结合具体的优选实施方式对本发明所作的进一步详细说明，不能认定本发明的具体实施只局限于这些说明。对于本发明所属技术领域的普通技术人员来说，在不脱离本发明构思的前提下，还可以做出若干简单推演或替换，都应当视为属于本发明的保护范围。

Claims (10)

  1. 一种用于可穿戴设备的新型身份验证方法，其特征在于：所述方法包括以下步骤：
    S1、可穿戴设备在用户敲击手背时，记录内置于可穿戴设备中加速度计和陀螺仪所检测到的三轴加速度数据和三轴角速度数据；
    S2、对采集到的加速度信号和角速度信号进行滤波降噪和端点切断处理；
    S3、对端点切断后的加速度信号和角速度信号进行对齐处理；
    S4、将对齐后的加速度数据和角速度数据进行快速傅里叶变换得到频域上的数据，拼接在原来的数据上；将处理完的信号数据作为训练数据集储存在可穿戴设备的数据库里；
    S5、用户敲击新的解锁信号数据作为测试数据，先检验测试数据是否为一次有效的敲击，然后用处理训练数据的方法来处理测试数据；
    S6、处理完的测试数据可以与训练集的数据通过机器学习分类方法进行分类，从而进行身份验证的识别。
  2. 根据权利要求1所述的方法，其特征在于，在步骤S1中，可穿戴设备的使用者敲击的是佩戴可穿戴设备的那只手；所述的敲击动作是指另外一只手手指敲击手背短暂而快速的运动。
  3. 根据权利要求1所述的方法，其特征在于，所述步骤S2包括：
    S21、分别对所述加速度计和陀螺仪检测到的数据进行滤波处理，获得加速度数据和角速度数据；
    S22、将获得滤波后的加速度数据和角速度数据进行切片处理，只取出等长的峰值及其附近的加速度数据和角速度数据，去掉没有振动信号的数据。
  4. 根据权利要求3所述的方法，其特征在于，所述步骤S3包括：
    S31、通过总体互相关法对端点切段后的振动信号进行对齐处理，所述对齐处理的具体操作是计算两个振动信号之间的偏移量，然后对当前的振动信号进行移动，移动完之后只取两个振动信号之间共有的完整部分；
    S32、对齐处理通过公式
    Figure PCTCN2018105076-appb-100001
    以及O(A,B)＝P(A,B)-n计算两个振动信号之间的偏移量O(A,B)，其中， a和b代表两个信号长度为n的振动信号，a(i)表示振动信号a的第i个点的振幅大小，b(i)表示振动信号b的第i个点的振幅大小，C(a,b)表示振动信号a和振动信号b的相关度；A表示对振动信号a两边长度为n的部分进行补零，进而获得的一个长度为3n的第一信号；B表示长度n的振动信号b；P(A,B)表示第一信号A中与第二信号B相关度最高的长度为n的信号位置；O(A,B)为计算所得的第一信号A与第二信号B之间的偏移量。
  5. 根据权利要求4所述的方法，其特征在于，所述步骤S4中，提取对齐处理后的振动信号的功率谱密度特征，并将所述功率谱密度特征与对齐处理前振动信号的振幅特征共同作为提取的信号特征。
  6. 根据权利要求5所述的方法，其特征在于，所述步骤S4中，通过公式
    Figure PCTCN2018105076-appb-100002
    提取对齐处理后的振动信号的功率谱密度特征PSD，其中，f s为振动信号的采样频率，n为信号长度，k表示信号长度为n的信号，FFT(k)表示对信号k的傅里叶变换，abs(FFT(k))表示对FFT(k)取绝对值。
  7. 根据权利要求1所述的方法，其特征在于，所述步骤S5包括：
    S51、用户敲击新的解锁信号数据作为测试数据，滤波降噪后的测试数据是否为一次有效的敲击，有效的敲击信号应该先是一段平缓的信号，接着是信号峰，最后是一段平缓的信号。只要检测信号前后若干个信号点的值是否超过设定的阈值，如果超过阈值，则丢弃该次采集的信号；
    S52、用处理训练数据的方法来处理测试数据，也就是切段，端点对齐处理，然后与训练数据进行对齐处理，最后快速傅里叶变换和计算功率谱密度计算。
  8. 根据权利要求1所述的方法，其特征在于，所述步骤S6包括：
    S61、多键组合身份验证：用户敲击除了大拇指外的4手指的关节骨的加速度数据和角速度数据作为训练数据，处理后将其保存在数据库里。并且将多键敲击的位置顺序作为组合密码也一并保存起来；
    S62、用户进行身份验证时的，每一次的敲击信号与数据库原有的所有敲击信号通过机器学习算法KNN进行对比分类，从而确定用户每一次敲击所对应的手背键位。多个键敲击组合密码与数据库原密码对比，从而进行验证；
    S63、单键敲击身份验证：用户只需重复敲击手背某一个位置作为训练数据保存在数据库里；
    S64、用户进行身份验证时，只需敲击手背一下，所产生的敲击信号与数据库保存的训练数据进行比较，从而进行验证。
  9. 根据权利要求1-3任一项所述的方法，其特征在于，所述步骤S2中：采用巴特沃兹滤波器对采集的振动信号进行滤波降噪处理，使用截止频率为20Hz的高通滤波滤除直流分量和低频噪音，使用截止频率为300Hz的低通滤波滤除高频噪音。
  10. 根据权利要求1-3任一项所述的方法，其特征在于：所述步骤S2中：所述端点切段处理中，使用固定长度来遍历整一段振动信号，当该段信号的能量最大时，则认为敲击信号出现，取该段长度前后一定长度的信号作为端点切段后的振动信号。
PCT/CN2018/105076 2018-08-31 2018-11-24 一种用于可穿戴设备的新型身份验证方法 WO2020042219A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811012680.3A CN109508728B (zh) 2018-08-31 2018-08-31 一种用于可穿戴设备的新型身份验证方法
CN201811012680.3 2018-08-31

Publications (1)

Publication Number Publication Date
WO2020042219A1 true WO2020042219A1 (zh) 2020-03-05

Family

ID=65745624

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/105076 WO2020042219A1 (zh) 2018-08-31 2018-11-24 一种用于可穿戴设备的新型身份验证方法

Country Status (2)

Country Link
CN (1) CN109508728B (zh)
WO (1) WO2020042219A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112990261A (zh) * 2021-02-05 2021-06-18 清华大学深圳国际研究生院 一种基于敲击节奏的智能手表用户识别方法
CN116127365A (zh) * 2023-04-14 2023-05-16 山东大学 一种基于振动的应用于智能头戴式设备的认证方法

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110321689A (zh) * 2019-07-08 2019-10-11 深圳大学 一种基于响指的身份识别方法和系统
CN110414196B (zh) * 2019-07-29 2021-09-17 深圳大学 一种基于振动信号的智能手表身份验证方法
CN110415415A (zh) * 2019-08-02 2019-11-05 深圳大学 一种用于门禁系统的身份验证方法
CN110968857B (zh) * 2019-12-03 2022-04-08 南京航空航天大学 基于抬臂动作的智能手表身份认证方法
CN112933579B (zh) * 2019-12-11 2023-03-31 中移(苏州)软件技术有限公司 一种运动质量评估方法及装置、存储介质
CN111585761B (zh) * 2020-03-30 2023-06-20 中国平安财产保险股份有限公司 基于特征函数的身份验证方法、装置和计算机设备
CN116055622B (zh) * 2022-06-24 2023-10-20 荣耀终端有限公司 一种功能触发方法、存储介质和终端
CN117272274A (zh) * 2023-10-08 2023-12-22 中国人民解放军总医院 一种智能电子保险柜及其身份验证方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103514387A (zh) * 2012-11-29 2014-01-15 Tcl集团股份有限公司 一种提高电子设备用户身份识别精度的方法及电子设备
CN105286845A (zh) * 2015-11-29 2016-02-03 浙江师范大学 一种适用于可穿戴式心率测量设备的运动噪声消除方法
CN105809438A (zh) * 2016-03-17 2016-07-27 广东乐心医疗电子股份有限公司 用于智能可穿戴设备的身份识别方法、支付方法与装置
CN105929940A (zh) * 2016-04-13 2016-09-07 哈尔滨工业大学深圳研究生院 基于细分特征值法的快速三维动态手势识别方法及系统
CN106454723A (zh) * 2016-09-09 2017-02-22 常州大学 一种基于手机加速度计的儿童监护方法

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105426814A (zh) * 2015-10-30 2016-03-23 苏州领泰数字医疗设备有限公司 一种基于手机的老人跌倒检测方法
CN105867656B (zh) * 2016-04-27 2019-03-08 牛亚锋 一种基于体感音乐穿戴设备的演奏动作识别方法
CN106406529A (zh) * 2016-09-19 2017-02-15 百度在线网络技术（北京）有限公司 一种输入方法、装置及可穿戴键盘
CN106921500B (zh) * 2017-03-22 2020-06-12 深圳先进技术研究院 一种移动设备的身份认证方法及装置
CN107426397A (zh) * 2017-04-18 2017-12-01 中国科学院计算技术研究所 基于用户行为特征的模型训练方法及身份验证方法
CN107153780B (zh) * 2017-05-05 2020-06-09 西安交通大学苏州研究院 基于手腕穿戴电子设备的书写行为特征认证方法
CN107037718A (zh) * 2017-05-25 2017-08-11 珠海市魅族科技有限公司 一种手表功能显示的控制方法及手表

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103514387A (zh) * 2012-11-29 2014-01-15 Tcl集团股份有限公司 一种提高电子设备用户身份识别精度的方法及电子设备
CN105286845A (zh) * 2015-11-29 2016-02-03 浙江师范大学 一种适用于可穿戴式心率测量设备的运动噪声消除方法
CN105809438A (zh) * 2016-03-17 2016-07-27 广东乐心医疗电子股份有限公司 用于智能可穿戴设备的身份识别方法、支付方法与装置
CN105929940A (zh) * 2016-04-13 2016-09-07 哈尔滨工业大学深圳研究生院 基于细分特征值法的快速三维动态手势识别方法及系统
CN106454723A (zh) * 2016-09-09 2017-02-22 常州大学 一种基于手机加速度计的儿童监护方法

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112990261A (zh) * 2021-02-05 2021-06-18 清华大学深圳国际研究生院 一种基于敲击节奏的智能手表用户识别方法
CN112990261B (zh) * 2021-02-05 2023-06-09 清华大学深圳国际研究生院 一种基于敲击节奏的智能手表用户识别方法
CN116127365A (zh) * 2023-04-14 2023-05-16 山东大学 一种基于振动的应用于智能头戴式设备的认证方法
CN116127365B (zh) * 2023-04-14 2023-07-25 山东大学 一种基于振动的应用于智能头戴式设备的认证方法

Also Published As

Publication number Publication date
CN109508728B (zh) 2020-08-25
CN109508728A (zh) 2019-03-22

Similar Documents

Publication Publication Date Title
WO2020042219A1 (zh) 一种用于可穿戴设备的新型身份验证方法
Wang et al. User authentication on mobile devices: Approaches, threats and trends
Chen et al. Taprint: Secure text input for commodity smart wristbands
Li et al. When CSI meets public WiFi: Inferring your mobile phone password via WiFi signals
Li et al. Whose move is it anyway? Authenticating smart wearable devices using unique head movement patterns
CN103917278B (zh) 传感器装置、分析装置与记录介质
CN105426713B (zh) 用于基于触摸事件分析来区分触摸屏使用者的方法和装置
Yu et al. Writinghacker: audio based eavesdropping of handwriting via mobile devices
Wu et al. Liveness is not enough: Enhancing fingerprint authentication with behavioral biometrics to defeat puppet attacks
Li et al. Velody: Nonlinear vibration challenge-response for resilient user authentication
Al-Naffakh et al. Unobtrusive gait recognition using smartwatches
Cheng et al. Magattack: Guessing application launching and operation via smartphone
Su et al. Towards device independent eavesdropping on telephone conversations with built-in accelerometer
Shen et al. WiPass: 1D-CNN-based smartphone keystroke recognition using WiFi signals
Cao et al. Evidence in hand: Passive vibration response-based continuous user authentication
Yang et al. EchoLock: Towards low-effort mobile user identification leveraging structure-borne echos
Wu et al. Power saving and secure text input for commodity smart watches
Wu et al. Toward robust detection of puppet attacks via characterizing fingertip-touch behaviors
Poletkin et al. A touch interface exploiting the use of vibration theories and infinite impulse response filter modeling based localization algorithm
Shen et al. Waving gesture analysis for user authentication in the mobile environment
Wei et al. SVSV: Online handwritten signature verification based on sound and vibration
Jiokeng et al. I Want to Know Your Hand: Authentication on Commodity Mobile Phones Based on Your Hand's Vibrations
Song et al. Pistis: Replay attack and liveness detection for gait-based user authentication system on wearable devices using vibration
Yu et al. Mobile devices based eavesdropping of handwriting
CN110414196B (zh) 一种基于振动信号的智能手表身份验证方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18932213

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 10/06/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 18932213

Country of ref document: EP

Kind code of ref document: A1