WO2020140266A1 - 一种智能手表的交互方法及交互系统 - Google Patents

一种智能手表的交互方法及交互系统 Download PDF

Info

Publication number
WO2020140266A1
WO2020140266A1 PCT/CN2019/070413 CN2019070413W WO2020140266A1 WO 2020140266 A1 WO2020140266 A1 WO 2020140266A1 CN 2019070413 W CN2019070413 W CN 2019070413W WO 2020140266 A1 WO2020140266 A1 WO 2020140266A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
algorithm
smart watch
vibration signals
interaction method
Prior art date
Application number
PCT/CN2019/070413
Other languages
English (en)
French (fr)
Inventor
伍楷舜
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
Priority to JP2019552836A priority Critical patent/JP2022518301A/ja
Priority to PCT/CN2019/070413 priority patent/WO2020140266A1/zh
Priority to US16/731,042 priority patent/US11270109B2/en
Publication of WO2020140266A1 publication Critical patent/WO2020140266A1/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
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
    • 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
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • 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
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Definitions

  • the invention relates to the field of interaction modes of smart devices, in particular to an interaction method and interaction system based on a smart watch.
  • wearable smart sensing devices are developing rapidly.
  • smart watches are especially popular, but because they are worn on the wrist, they cannot be equipped with a large enough screen, and people cannot input as they like on a mobile phone.
  • the existing input methods of smart watches are mainly divided into three types: single-touch screen, finger tracking and voice recognition.
  • the methods of single-touch screen and finger tracking are limited by the screen, while voice recognition is more limited due to the sensitivity of information.
  • many research teams have also conducted research, most of which require additional equipment, and the existence of purchase costs and learning costs, most of which cannot be widely accepted.
  • the present invention proposes a smart watch interaction method and interaction system based on human body propagating vibration signals, and as far as possible fits the user's usage habits, develop new interaction methods for smart watches, thereby solving the inferiority of smart watch interaction methods The problem.
  • the present invention adopts the following technical solutions:
  • a smart watch interaction method includes the following steps:
  • the X, Y, and Z axis vibration signals of the accelerometer and the gyroscope are collected separately.
  • step S2 using the anomaly detection algorithm to identify the signal includes:
  • step S3 includes:
  • step S32 In the stage of initializing the training model, the data processed in step S31 is stored as a training sample in the database; in the actual use stage, the improved algorithm based on the k-nearest neighbor algorithm is used to classify and identify the signal.
  • the improved algorithm of the K-neighbor algorithm is specifically: based on the dynamic time warping algorithm, the actual signal and the training signal are matched in units of frames, and the shortest Manhattan distance between each other is calculated and used as the k-nearest neighbor The basis for classification and recognition by the algorithm.
  • step S4 includes:
  • the training sample is corrected to a certain degree, so as to maintain the stability of the accuracy.
  • step S41 actual input is corrected by providing candidate keys or by associating results in the input method.
  • step S42 is specifically:
  • An interactive system for smart watches including:
  • the signal detection module collects the vibration signals of the accelerometer and gyroscope of the smart watch based on the vibration signals transmitted by the human body;
  • Recognition and classification module using abnormal detection algorithm to identify vibration signals; preprocessing the vibration signals, and using k-nearest neighbor algorithm to improve the vibration signals for further classification and recognition;
  • the real-time feedback module analyzes user feedback on the results and corrects them in time to maintain stable recognition accuracy.
  • a program that executes the smart watch interaction method of the present invention is a program that executes the smart watch interaction method of the present invention.
  • the beneficial effect of the present invention lies in: by using human body parts (for example, the back of the hand) as a virtual screen based on the propagation of vibration signals of the human body, combined with an improved machine learning algorithm, the interactive means of the watch is actually expanded The user experience is improved.
  • the interactive method of the present invention is novel and interesting, can effectively meet user needs, and can be widely applied to text input, watch games, and the like.
  • FIG. 6 is the matching result of the dynamic time warping algorithm in frame units after the improvement of the present invention (frame shift is 1, frame length is 3);
  • the invention discloses a smart watch interaction method and interaction system based on human body propagating vibration signals and machine learning. As shown in FIG. 1, the interactive method of the present invention includes the following steps:
  • the program controls the accelerometer and gyroscope of the smart watch to collect the vibration signals of the smart watch accelerometer and the gyroscope;
  • the vibration signals of the X, Y, and Z axes of the accelerometer and the gyroscope are collected separately;
  • step S2 uses an anomaly detection algorithm to identify the signal.
  • the specific steps are as follows:
  • the end position of the signal When the continuous amplitude is less than the noise signal threshold, set the end position of the signal to the current position, preferably the signal length is 10 points, the noise signal threshold Is 0.015. After detecting the signal segment between the noise signal threshold and the effective knock signal threshold, the constraints of the signal length and the signal-to-noise ratio are added. Preferably, the signal length L satisfies 37 ⁇ L ⁇ 60.
  • the signal length After the signal length satisfies the constraint conditions, it is calculated separately After filtering the energy of m points before the signal and the energy after m points, determine whether the signal is greater than the signal-to-noise ratio threshold, if it is greater, the signal is judged to be a valid signal, otherwise it is considered to be a noise signal, and the signal-to-noise ratio threshold is S10. Can be detected.
  • step S3 pre-processes the vibration signal, and uses the improved algorithm of k-nearest neighbor algorithm to further classify the vibration signal.
  • the specific steps are as follows:
  • the vibration signals of the X, Y, and Z axes of the accelerometer and the gyroscope of each sample are spliced by sensor type, and the 3-axis data of the corresponding sensor is normalized as a whole, and the average value of the data is subtracted.
  • the normalized data is stored as a training sample in the database, and in the actual use stage, the signal is classified and recognized based on the improved algorithm of the k-nearest neighbor algorithm. Specifically, based on the dynamic time warping algorithm, the distance between the test/input sample and the training sample is calculated, and the classification result is given according to the size of the distance.
  • the dynamic time warping algorithm is a kind of thinking based on dynamic programming.
  • the object of dynamic time warping is expanded from the original one-dimensional point to the three-dimensional (three-axis) frame, and the distance between each other is calculated, which can be more accurate
  • the algorithm allows the frame length and frame shift to be adjusted according to the actual sampling frequency and demand to reduce the power consumption of the algorithm, thereby obtaining the ideal performance.
  • the distance is not limited to Manhattan distance or Euler distance.
  • Fig. 5 is the signal matching result of the original dynamic time warping algorithm
  • Fig. 6 is the matching result of the dynamic time warping algorithm in units of frames after the improvement of the present invention (frame shift is 1, frame length is 3); After the time warping algorithm increases the constraints of frame length and frame shift, the signal matching method will indeed change.
  • step S4 of this embodiment after obtaining the classification result in step S2, the result is output to the application, and at the same time, the new sample X and the training sample obtained by the S3 algorithm The distance is recorded and the application feedback is monitored. After receiving feedback on the classification results, the training samples are operated according to the established sample replacement strategy, thereby obtaining higher robustness. Specifically: after the classification result set obtained in step 3, the user input results are corrected, specifically by correcting the actual input by providing candidate keys or by associating results in the input method; after correction, the training sample is corrected to a certain extent, Therefore, the accuracy is kept stable.
  • the correction result is consistent with the classification result, and there is no operation; when the correction result is inconsistent with the classification result, for the samples of the same category as the classification result in the training sample, delete the algorithm calculated by the improved algorithm of the K-neighbor algorithm The sample with the largest distance, and then replace the current sample to the position of the deleted sample.
  • the structure of the specific implementation of this example is divided into three modules, a signal detection module, a recognition classification module, and a real-time feedback module.
  • the signal detection module detects the signal, normalizes the signal after detecting the signal, subtracts the mean of the data and divides by the variance of the data, as the input of the classification recognition module; the training (initialization) stage of the classification recognition module is pure The signal storage operation, and put into use after the training is completed, will perform an improved classification algorithm.
  • the classification results will be passed into the real-time feedback module.

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • User Interface Of Digital Computer (AREA)

Abstract

一种智能手表的交互方法，包括以下步骤：（S1）基于人体传播振动信号，采集智能手表加速度计和陀螺仪的振动信号；（S2）采用异常检测算法识别振动信号；（S3）对振动信号进行预处理，采用k近邻算法改进后的算法对振动信号进行进一步的分类识别；（S4）分析使用者对结果的反馈，及时矫正以维持稳定的识别精度。还公开了其交互系统，包括信号检测模块、识别分类模块、实时反馈模块。通过基于人体传播振动信号把人体部位当作虚拟屏幕，并结合了改进的机器学习算法，实际地拓展了手表的交互手段，提高了用户体验，交互手段新颖有趣，能够切实满足用户需求，可广泛应用于文本输入、手表游戏等。

Description

一种智能手表的交互方法及交互系统 技术领域
本发明涉及智能设备的交互方式领域，具体涉及基于一种智能手表的交互方法及交互系统。
背景技术
目前，可穿戴智能感知设备迅速发展，其中智能手表尤为流行，但由于其佩戴于手腕，无法配备足够大的屏幕，人们无法像手机一样地随心所欲地输入。智能手表现有的输入方式主要分为三种：单点触屏、手指跟踪以及语音识别。单点触屏以及手指跟踪的方法受屏幕的限制，而语音识别则由于信息敏感更是颇受限制。为了摆脱智能手表输入困难的窘境，如今许多科研团队也进行了研究，大多需要额外配置设备，购置成本以及学习成本的存在，大多无法广泛地被接受。
发明内容
针对上述技术问题，本发明提出了基于人体传播振动信号的智能手表的交互方法及交互系统，尽量贴合用户使用习惯的情况下，为智能手表拓展新的交互手段，从而解决智能手表交互手段贫瘠的问题。本发明采用了如下技术方案：
一种智能手表的交互方法，包括以下步骤：
S1、基于人体传播振动信号，采集智能手表加速度计和陀螺仪的振动信号；
S2、采用异常检测算法识别振动信号；
S3、对振动信号进行预处理，采用k近邻算法改进后的算法对振动信号进行进一步的分类识别；
S4、分析使用者对结果的反馈，及时矫正以维持稳定的识别精度。
进一步，分别采集加速度计和陀螺仪的X、Y、Z三轴的振动信号。
进一步，步骤S2利用异常检测算法来识别信号包括：
S21、采集加速度计Z轴数据；
S22、采用高通滤波器对所述加速度计Z轴数据进行滤波；
S23、设置有效敲击信号的阈值及噪音信号阈值；
S24、读取一段振幅小于噪音信号阈值信号作为第一状态；
S25、继续监听，等待读取振幅大于有效敲击信号阈值的信号，计振幅大于有效敲击信号阈值的位置为x，设置信号的起始位置为位置X前L的位置，即X-L；
S26、继续监听，等待读取一段连续且振幅小于噪音信号阈值的信号，当出现连续振幅小于噪音信号阈值时，设置信号的结束位置为当前位置；
S27、通过信号起止位置获取信号数据，判断信号长度是否满足长度区间，如不满足则回到S25，如满足则进入下一步；
S28、对数据进行高通滤波，分别计算滤波后信号前m个点的能量及m个点之后的能量，判断信号是否大于信噪比阈值，大于则判断信号为有效信号，否则认为是噪音信号，返回到S25。
进一步，步骤S3包括：
S31、对信号进行归一化的预处理，将信号减去均值并除以方差；
S32、在初始化训练模型阶段，将步骤S31处理后的数据作为训练样本存储到数据库中；在实际使用阶段，使用基于k近邻算法改进后的算法对信号进行分类与识别。
进一步，K邻近算法改进后的算法具体为：基于动态时间规整算法，将实际信号与训练信号进行以帧为单位的匹配，计算出彼此之间的最短的曼哈顿距离，并以此来作为k近邻算法进行分类识别的依据。
进一步，步骤S4包括：
S41、由步骤3得到的分类结果集合后，纠正使用者输入结果；
S42、纠正后，对训练样本进行一定程度的修正，从而保持精度的稳定。
进一步，步骤S41中通过提供候选键或通过在输入法的联想结果对于实际输入进行纠正。
进一步，步骤S42具体为：
S421、纠正结果与分类结果一致，无操作；
S422、纠正结果与分类结果不一致时，对于训练样本中与分类结果相同类别的样本，删除通过K邻近算法改进后的算法计算得到的距离最大的样本，再将当前的样本替换至被删除样本的位置。
一种智能手表的交互系统，包括：
信号检测模块，基于人体传播振动信号，采集智能手表加速度计和陀螺仪的振动信号；
识别分类模块，采用异常检测算法识别振动信号；对振动信号进行预处理，采用k近邻算法改进后的算法对振动信号进行进一步的分类识别；
实时反馈模块，分析使用者对结果的反馈，及时矫正以维持稳定的识别精度。
一种程序，执行本发明所述的智能手表的交互方法。
与现有技术相比，本发明的有益效果在于：通过基于人体传播振动信号把人体部位(例如，手背)当作虚拟屏幕，并结合了改进的机器学习算法，实际地拓展了手表的交互手段，提高了用户体验，本发明的交互手段新颖有趣，能够切实满足用户需求，可广的应用于文本输入、手表游戏等。
附图说明
图1是本发明的交互方法的流程图；
图2是本发明的检测信号模块的工作流程图；
图3是本发明的反馈系统工作流程图；
图4为本发明的结构示意图；
图5是本发明原始动态时间规整算法的信号匹配结果；
图6是本发明改进后以帧为单位进行动态时间规整算法的匹配结果(帧移为1，帧长 为3)；
具体实施方式
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。
下面结合附图对本发明的较优的实施例作进一步的详细说明。
本发明公开了基于人体传播振动信号与机器学习的智能手表交互方法及交互系统。如图1所述，本发明的交互方法包括了以下步骤：
S1、基于人体传播振动信号，程序控制智能手表的加速度计和陀螺仪采集智能手表加速度计和陀螺仪的振动信号；
具体为分别采集加速度计和陀螺仪的X、Y、Z三轴的振动信号；
S2、采用异常检测算法识别振动信号；
S3、对振动信号进行预处理，采用k近邻算法改进后的算法对振动信号进行进一步的分类识别；
S4、分析使用者对结果的反馈，及时矫正以维持稳定的识别精度。
如图2所示，步骤S2利用异常检测算法来识别信号，具体步骤如下：
先采集加速度计Z轴数据，并取高通40hz的信号，同时读取一段振幅小于噪声信号阈值的信号，优选信号长度为10个点，噪声信号阈值为0.015，此时为信号检测的第一个状态；继续监听，等待振幅大于有效敲击信号阈值的信号，计振幅大于有效敲击信号阈值的位置为x，设置信号的起始位置为位置X前L的位置，即X-L，优选有效敲击信号阈值为2。在接收目标信号后，等待读取一段连续的振幅小于噪声信号阈值的信号，当出现连续振幅小于噪音信号阈值时，设置信号的结束位置为当前位置，优选信号长度为10个点，噪声信号阈值为0.015。在噪声信号阈值与有效敲击信号阈值之间检测得到的信号片段之后，再增加信号长度与信噪比的约束，优选信号长度L满足37≤L≤60，信号长度满足约束条件后，分别计算滤波后信号前m个点的能量及m个点之后的能量，判断信号是否大于信噪比阈值，大于则判断信号为有效信号，否则认为是噪音信号，信噪比阈值为S10，至此，信号得以被检测出来。
本实施例中，步骤S3对振动信号进行预处理，采用k近邻算法改进后的算法对振动信号进行进一步的分类识别，具体步骤如下：
先对每一个样本的加速度计和陀螺仪的X、Y、Z三轴的振动信号以传感器类别划分的拼接，针对对应传感器的3轴数据进行整体的归一化处理，具体减去数据的均值，再除以数据的方差，然后在初始化训练模型阶段，将归一化处理后的数据作为训练样本存储到数据库中，在实际使用阶段基于k近邻算法改进后的算法对信号进行分类与识别，具体为基于动态时间规整算法，计算测试/输入样本与训练样本之间的距离，根据距离的大小给出分类结果。
其中，动态时间规整算法，是一种基于动态规划的思路，将动态时间规整的对象由从原始的一维的点扩展为三维(三轴)的帧，并计算彼此的距离，可以更准确地衡量两个信号的相似程度，同时，算法允许根据实际的采样频率与需求相应地调整帧长与帧移来降低算法的功耗，从而获得理想的性能。解决由于两个信号之间的时序错位而无法进行比较的问题，并将两者之间的区别进行量化的方法。所述的距离不限于为曼哈顿距离或欧拉距离。
图5是原始动态时间规整算法的信号匹配结果；图6是本发明改进后以帧为单位进行动态时间规整算法的匹配结果(帧移为1，帧长为3)；由此可看出动态时间规整算法在增加了帧长帧移的约束之后，信号匹配方式确实会发生改变。
如图3所示，本实施例的步骤S4中，在步骤S2获得分类结果之后，将结果输出至应用中，并且与此同时，将新的样本X以及S3算法中所得到的与训练样本的距离记录下来，监听应用的反馈。在接收到对分类结果的反馈后，根据既定的样本替换策略来对训练样本进行操作，从而获得较高的鲁棒性。具体为：由步骤3得到的分类结果集合后，纠正使用者输入结果，具体通过提供候选键或通过在输入法的联想结果对于实际输入进行纠正；纠正后，对训练样本进行一定程度的修正，从而保持精度的稳定，具体为纠正结果与分类结果一致，无操作；纠正结果与分类结果不一致时，对于训练样本中与分类结果相同类别的样本，删除通过K邻近算法改进后的算法计算得到的距离最大的样本，再将当前的样本替换至被删除样本的位置。
如图4所示，本例的具体实施的结构分为三个模块，信号检测模块、识别分类模块，以及实时反馈模块。信号检测模块检测信号，检测到信号后将信号进行归一化预处理，减去数据的均值再除以数据的方差，作为分类识别模块的输入；分类识别模块的训练(初始化)阶段为单纯的信号存储操作，而在训练完成后投入使用阶段，将执行改进的分类算法。分类结果将传递进入实时反馈模块。
以上内容是结合具体/优选的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，其还可以对这些已描述的实施方式做出若干替代或变型，而这些替代或变型方式都应当视为属于本发明的保护范围。

Claims (11)

  1. 一种智能手表的交互方法，其特征在于，包括以下步骤：
    S1、基于人体传播振动信号，采集智能手表加速度计和陀螺仪的振动信号；
    S2、采用异常检测算法识别振动信号；
    S3、对振动信号进行预处理，采用k近邻算法改进后的算法对振动信号进行进一步的分类识别；
    S4、分析使用者对结果的反馈，及时矫正以维持稳定的识别精度。
  2. 根据权利要求1所述的智能手表的新型交互方法，其特征在于，分别采集加速度计和陀螺仪的X、Y、Z三轴的振动信号。
  3. 根据权利要求2所述的智能手表的新型交互方法，其特征在于，步骤S2利用异常检测算法来识别信号包括：
    S21、采集加速度计Z轴数据；
    S22、采用高通滤波器对所述加速度计Z轴数据进行滤波；
    S23、设置有效敲击信号的阈值及噪音信号阈值；
    S24、读取一段振幅小于噪音信号阈值信号作为第一状态；
    S25、继续监听，等待读取振幅大于有效敲击信号阈值的信号，计振幅大于有效敲击信号阈值的位置为x，设置信号的起始位置为位置X前L的位置，即X-L；
    S26、继续监听，等待读取一段连续且振幅小于噪音信号阈值的信号，当出现连续振幅小于噪音信号阈值时，设置信号的结束位置为当前位置；
    S27、通过信号起止位置获取信号数据，判断信号长度是否满足长度区间，如不满足则回到S25，如满足则进入下一步；
    S28、对数据进行高通滤波，分别计算滤波后信号前m个点的能量及m个点之后的能量，判断信号是否大于信噪比阈值，大于则判断信号为有效信号，否则认为是噪音信号，返回到S25。
  4. 根据权利要求1所述的智能手表的新型交互方法，其特征在于，步骤S3具体为：
    S31、对信号进行归一化的预处理，将信号减去均值并除以方差；
    S32、在初始化训练模型阶段，将步骤S31处理后的数据作为训练样本存储到数据库中；在实际使用阶段，使用基于k近邻算法改进后的算法对信号进行分类与识别。
  5. 根据权利要求1所述的智能手表的新型交互方法，其特征在于，K邻近算 法改进后的算法具体为：基于动态时间规整算法，将实际信号与训练信号进行以帧为单位的匹配，计算出彼此之间的最短的距离，并以此来作为k近邻算法进行分类识别的依据。
  6. 根据权利要求5所述的智能手表的新型交互方法，其特征在于，所述距离为曼哈顿距离或欧拉距离。
  7. 根据权利要求1所述的智能手表的新型交互方法，其特征在于，步骤S4包括：
    S41、由步骤3得到的分类结果集合后，纠正使用者输入结果；
    S42、纠正后，对训练样本进行一定程度的修正，从而保持精度的稳定。
  8. 根据权利要求7中所述的智能手表的新型交互方法，其特征在于，步骤S41中通过提供候选键或通过在输入法的联想结果对于实际输入进行纠正。
  9. 根据权利要求8中所述的智能手表的新型交互方法，其特征在于，步骤S42具体为：
    S421、纠正结果与分类结果一致，无操作；
    S422、纠正结果与分类结果不一致时，对于训练样本中与分类结果相同类别的样本，删除通过K邻近算法改进后的算法计算得到的距离最大的样本，再将当前的样本替换至被删除样本的位置。
  10. 一种智能手表的交互系统，其特征在于，包括：
    信号检测模块，基于人体传播振动信号，采集智能手表加速度计和陀螺仪的振动信号；
    识别分类模块，采用异常检测算法识别振动信号；对振动信号进行预处理，采用k近邻算法改进后的算法对振动信号进行进一步的分类识别；
    实时反馈模块，分析使用者对结果的反馈，及时矫正以维持稳定的识别精度。
  11. 一种程序，执行权利要求1-9任一项所述的智能手表的交互方法。
PCT/CN2019/070413 2019-01-04 2019-01-04 一种智能手表的交互方法及交互系统 WO2020140266A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
JP2019552836A JP2022518301A (ja) 2019-01-04 2019-01-04 スマートウォッチのヒューマン・マシンインタラクション方法並びにそのヒューマン・マシンインタラクションシステム及びプログラム
PCT/CN2019/070413 WO2020140266A1 (zh) 2019-01-04 2019-01-04 一种智能手表的交互方法及交互系统
US16/731,042 US11270109B2 (en) 2019-01-04 2019-12-31 Interactive method and interactive system for smart watch

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/070413 WO2020140266A1 (zh) 2019-01-04 2019-01-04 一种智能手表的交互方法及交互系统

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/731,042 Continuation-In-Part US11270109B2 (en) 2019-01-04 2019-12-31 Interactive method and interactive system for smart watch

Publications (1)

Publication Number Publication Date
WO2020140266A1 true WO2020140266A1 (zh) 2020-07-09

Family

ID=71407141

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/070413 WO2020140266A1 (zh) 2019-01-04 2019-01-04 一种智能手表的交互方法及交互系统

Country Status (3)

Country Link
US (1) US11270109B2 (zh)
JP (1) JP2022518301A (zh)
WO (1) WO2020140266A1 (zh)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120319940A1 (en) * 2011-06-16 2012-12-20 Daniel Bress Wearable Digital Input Device for Multipoint Free Space Data Collection and Analysis
CN103543826A (zh) * 2013-07-30 2014-01-29 广东工业大学 一种基于加速度传感器的手势识别方法
CN105353881A (zh) * 2015-12-04 2016-02-24 深圳大学 基于rfid的手势识别方法及系统
CN106237604A (zh) * 2016-08-31 2016-12-21 歌尔股份有限公司 可穿戴设备及利用其监测运动状态的方法
CN106919958A (zh) * 2017-03-21 2017-07-04 电子科技大学 一种基于智能手表的人体手指动作识别方法
CN107329682A (zh) * 2017-06-19 2017-11-07 努比亚技术有限公司 边缘交互方法及移动终端
CN108292167A (zh) * 2015-11-30 2018-07-17 索尼公司 信息处理设备、信息处理方法和程序

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5962505B2 (ja) * 2010-03-15 2016-08-03 日本電気株式会社 入力装置、入力方法及びプログラム
US9168419B2 (en) * 2012-06-22 2015-10-27 Fitbit, Inc. Use of gyroscopes in personal fitness tracking devices
KR102384225B1 (ko) * 2015-03-06 2022-04-07 삼성전자주식회사 혈압 측정 장치 및 방법
WO2018005933A1 (en) * 2016-07-01 2018-01-04 Intel Corporation Technologies for user-assisted machine learning
US20180239435A1 (en) * 2017-02-22 2018-08-23 International Business Machines Corporation Smart devices having recognition features

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120319940A1 (en) * 2011-06-16 2012-12-20 Daniel Bress Wearable Digital Input Device for Multipoint Free Space Data Collection and Analysis
CN103543826A (zh) * 2013-07-30 2014-01-29 广东工业大学 一种基于加速度传感器的手势识别方法
CN108292167A (zh) * 2015-11-30 2018-07-17 索尼公司 信息处理设备、信息处理方法和程序
CN105353881A (zh) * 2015-12-04 2016-02-24 深圳大学 基于rfid的手势识别方法及系统
CN106237604A (zh) * 2016-08-31 2016-12-21 歌尔股份有限公司 可穿戴设备及利用其监测运动状态的方法
CN106919958A (zh) * 2017-03-21 2017-07-04 电子科技大学 一种基于智能手表的人体手指动作识别方法
CN107329682A (zh) * 2017-06-19 2017-11-07 努比亚技术有限公司 边缘交互方法及移动终端

Also Published As

Publication number Publication date
JP2022518301A (ja) 2022-03-15
US11270109B2 (en) 2022-03-08
US20200226369A1 (en) 2020-07-16

Similar Documents

Publication Publication Date Title
CN109840480B (zh) 一种智能手表的交互方法及交互系统
WO2018040757A1 (zh) 可穿戴设备及利用其监测运动状态的方法
US10339371B2 (en) Method for recognizing a human motion, method for recognizing a user action and smart terminal
US20120323521A1 (en) System and method for recognizing gestures
CN105184325B (zh) 一种移动智能终端
CN104135911B (zh) 多轴线活动监测装置中的活动分类
WO2017092225A1 (zh) 基于emg的可穿戴式文本输入系统及方法
WO2020042711A1 (zh) 一种提高睡眠监测准确性的可穿戴设备
US20150091790A1 (en) Classification of gesture detection systems through use of known and yet to be worn sensors
CN107440680A (zh) 睡眠状态推定装置
KR102036987B1 (ko) 생체 및 움직임 정보를 이용한 수면 상태 감지 방법 및 장치
US10620012B2 (en) Step counting method, device, and terminal
US10241583B2 (en) User command determination based on a vibration pattern
CN112057066A (zh) 心率的检测方法、可穿戴设备及计算机存储介质
CN109063545A (zh) 一种疲劳驾驶检测方法及装置
WO2017185222A1 (zh) 一种基于球类运动的运动轨迹采集和分析的系统与方法
CN109758154B (zh) 一种运动状态确定方法、装置、设备及存储介质
WO2020140266A1 (zh) 一种智能手表的交互方法及交互系统
WO2018014432A1 (zh) 语音应用触发控制方法、装置及终端
CN105138129A (zh) 左右手识别方法、装置及电子设备
CN107669282A (zh) 基于面部识别的测谎仪
US20180353125A1 (en) Method and apparatus for determining sleep state using biometric information and motion information
US20210282656A1 (en) Blood pressure monitoring method, apparatus, and device
CN111803902B (zh) 泳姿识别方法、装置、可穿戴设备及存储介质
CN104808778A (zh) 判断头戴式智能设备操作有效性的装置和方法

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2019552836

Country of ref document: JP

Kind code of ref document: A

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

Ref document number: 19907478

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 02.11.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19907478

Country of ref document: EP

Kind code of ref document: A1