WO2019080734A1 - 一种基于WiFi信号的情绪识别方法 - Google Patents

一种基于WiFi信号的情绪识别方法

Info

Publication number
WO2019080734A1
WO2019080734A1 PCT/CN2018/110206 CN2018110206W WO2019080734A1 WO 2019080734 A1 WO2019080734 A1 WO 2019080734A1 CN 2018110206 W CN2018110206 W CN 2018110206W WO 2019080734 A1 WO2019080734 A1 WO 2019080734A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
csi
emotions
extracting
fluctuation curve
Prior art date
Application number
PCT/CN2018/110206
Other languages
English (en)
French (fr)
Inventor
叶伟
Original Assignee
叶伟
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 叶伟 filed Critical 叶伟
Publication of WO2019080734A1 publication Critical patent/WO2019080734A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
    • A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/08—Testing, supervising or monitoring using real traffic
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
    • 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 belongs to the field of artificial intelligence technology, and in particular relates to an emotion recognition method based on a WiFi signal.
  • ECG electrocardiogram
  • the invention provides an emotion recognition method based on a WiFi signal, which does not require the user to wear any device, and only needs to be installed in an ordinary WiFi device such as a home or an office.
  • the CSI signal in WiFi is used to describe the weakening factors of the signal in each transmission path, such as signal scattering, environmental attenuation, distance attenuation and other information, which is highly sensitive to weak signals.
  • the invention extracts the user's strong positive emotion and negative emotion feature data through the CSI signal data in the WiFi, and recognizes the user's inherent true emotion.
  • An emotion recognition method based on a WiFi signal comprising the following steps:
  • A1 collecting wireless signals sent by the WiFi router under different emotions
  • A3 for CSI data processing, modeling, extracting ventricular fluctuation curve, atrial fluctuation curve and respiratory fluctuation curve;
  • step A6 according to the measured accuracy rate, return to step A4 for iterative analysis
  • the processing for CSI data includes:
  • B1 data preprocessing including:
  • the CSI data is normalized according to the influence of the transmission power of different WiFi devices on the CSI data
  • Multipath suppression that is, transforming CSI from the frequency domain to the time domain (IFFT), and then removing the multi-path changes caused by environmental changes and human movement by removing signal components of long delay that are not related to heart rate and breathing;
  • Wavelet filtering that is, using discrete wavelet transform to remove high frequency noise
  • step A4 includes:
  • Extracting the mood fluctuation curve that is, extracting and analyzing the periodicity data of the heart rate and respiratory related data by analyzing the characteristics of the CSI frequency domain data; and matching and contrasting with the time, duration and duration of the emotion recorded in the laboratory.
  • IPI Inter-Peak-Interval, peak interval
  • SDIPI standard deviation of peak interval
  • the related feature data is converted into a vector, and as an input, a deep learning algorithm is used for classification and recognition, and the classification result is recorded, compared with the original data analysis, the precision rate and the recall rate data are analyzed, and the model is iterated.
  • the invention relates to the feature extraction of CSI (Channel State Information) signals in WiFi in the communication technology for different emotions, and the real emotion of the user is recognized, which has higher accuracy than the computer vision emotion recognition.
  • CSI Channel State Information
  • the invention recognizes the user's strong emotion by analyzing the CSI signal in the WiFi through the indoor installed Wi-Fi router device, and can improve the emotional ability of the human-computer interaction in the smart home life, and truly recognize the user's internal intention. In addition, you can actively alleviate the user's current bad mood, avoiding various diseases caused by bad mood, such as stomach, heart disease and so on.
  • An advantage of the present invention is that the intrinsic emotion of the user is relatively accurately recognized using a conventional WiFi device.
  • the CSI signal has a relatively high multipath resolution capability, especially for small changes in signals in the non-line of sight range.
  • the CSI signal also has high sensitivity and perceptual breadth for emotion recognition.
  • User sentiment is related to heart rate changes, and user emotions tend to rise in heart rate and breathing. The process of reaching the cap and then attenuating, a curve of inverted U, is not the same as the periodic fluctuations of heart rate and breathing; There are very different characteristics between them, through clustering analysis and supervised classification and identification of these data through feature indicators.
  • the invention discriminates the user's "psychological activity" characteristic data through the CSI signal data, and accelerates the user's heartbeat acceleration, breathing acceleration, and the duration of the two, and recognizes the user's inherently strong emotions. The visually recognized emotions are more accurate;
  • Sensitivity Through the extraction of data features, atrial fluctuation curve, ventricular fluctuation curve, mood fluctuation curve, IBI (inter-beat-interval), SDIBI (standard interval between heart beats) Data information, and detailed analysis of the internal emotions of different characteristics of users;
  • FIG. 1 is a WiFi emotion recognition apparatus involved in an embodiment of the present invention.
  • Fig. 2 is a diagram showing the emotion classification involved in the present invention.
  • FIG. 3 is a flowchart of an emotion recognition method based on a WiFi signal in an embodiment of the present invention.
  • FIG. 4 is a schematic diagram of a process of emotion recognition in an embodiment of the present invention.
  • the 802.11n-based WiFi technology uses a MIMO-OFDM system that utilizes the software provided by Daniel Halperin of the University of Washington to acquire 30 subcarriers in a wireless communication channel. Only one Wi-Fi router is needed. A remote server is used to process emotion recognition, so you can view your emotion recognition on your mobile phone, computer and other devices. It is also possible to directly run the emotion recognition project on your mobile phone, on the PC, and view the recognition results in real time in a similar terminal device.
  • the Wi-Fi emotion recognition device is shown in Figure 1.
  • Wi-Fi technology based on 802.11n protocol uses MIMO-OFDM system (MIMO, Multiple-Input Multiple-Output, OFDM, Orthogonal Frequency Division Multiplexing, Orthogonal Frequency Division Multiplexing), using Daniel from the University of Washington
  • MIMO-OFDM system MIMO, Multiple-Input Multiple-Output, OFDM, Orthogonal Frequency Division Multiplexing, Orthogonal Frequency Division Multiplexing
  • the tool software provided by Halperin can acquire 30 subcarriers in a wireless communication channel; finally, the normalized CSI matrix H——n*m*30 complex matrix can be obtained, where n represents the number of transmitting antennas, and m represents the receiving antennas.
  • the number 30 is the number of subcarrier information.
  • a modular matrix of 2*3*30 channel information is as follows:
  • the classification of emotions in the present invention is based on a circular pattern of emotion classification by Russell (1980), which divides emotion into two dimensions: happiness and intensity.
  • the happiness is divided into pleasant and unpleasant, the intensity is divided into medium intensity and high intensity, and low-intensity emotion recognition is not considered.
  • the eight emotions in the following figure are identified by the classification algorithm: surprise, happiness; satisfaction, fatigue; decadence, depression Angry, frightened.
  • the emotional data collection experiment design includes:
  • the experimenter first registers the personal information, the emotions are stable, and then presents the small cards into the room with the emotions designed in advance;
  • the hardware equipment is designed to be simple.
  • the difficulty lies in acquiring the heart rate fluctuation curve (atrial, ventricular-induced multipath effect) and respiratory fluctuation curve (multiple-path effect caused by the chest) related to mood fluctuations.
  • the happy, strongly feature-related data is extracted on the frequency domain data, and the classification is realized by a deep learning algorithm.
  • the process of model training mainly includes the following steps: 1. Accepting indoor users to send signals to WiFi routers under different emotions, collecting data uploaded by WiFi; 2. Extracting CSI signals from wireless signals; 3. Processing CSI data, including data pre-processing Processing, de-noise, etc., and then on the periodic fluctuations in heart rate, respiratory periodic fluctuations to do data characteristics.
  • the method for modeling and correcting emotion recognition includes the following steps:
  • Heart rate changes caused by emotions, respiratory changes are even smaller changes, using bandpass filtering to filter high frequency data, using a 2nd order butterworth filter; on the other hand, because different Wi-Fi device transmit power has a certain impact on CSI data, need to The data is normalized;
  • Multipath suppression transforming CSI from the frequency domain to the time domain (IFFT), and then removing the multi-path changes caused by environmental changes and human movement by removing signal components of long delay that are not related to heart rate and breathing;
  • Wavelet filtering The high-frequency noise is removed by discrete wavelet transform, in which the threshholding part adopts dynamic threshold and 4 layers of Symlet wavelet;
  • the collected data will exclude some factors that are not related to emotions, individual differences data, such as gender, age, current activity, etc.
  • the heart rate and respiratory related channel data are extracted, and these periodic data are extracted; and the time and duration of the emotion recorded in the laboratory are matched and compared, and the F test is used statistically. Identify data with significant fluctuations;
  • Atrial fluctuation curves ventricular fluctuation curve, respiratory curve, analysis of IPI (Inter-Peak-Interval, peak interval), SDIPI, the standard deviation of peak interval, time window is 2 minutes; and analysis of power spectrum performance of different emotions Cluster analysis of these data, analysis and adjustment of relevant feature data. Then convert these feature data into a vector, use as a input, use the deep learning algorithm to classify and identify, and record the classification results, compare with the original data analysis, analyze the accuracy rate, recall rate and other data, and iterate the model.
  • IPI Inter-Peak-Interval, peak interval
  • SDIPI the standard deviation of peak interval

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Psychiatry (AREA)
  • Social Psychology (AREA)
  • Biomedical Technology (AREA)
  • Educational Technology (AREA)
  • Hospice & Palliative Care (AREA)
  • Psychology (AREA)
  • Child & Adolescent Psychology (AREA)
  • Electromagnetism (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Developmental Disabilities (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

一种基于WiFi信号的情绪识别方法，包括步骤：A1，收集在不同情绪下，WiFi路由器发出的无线信号；A2，从无线信号中提取出CSI信号；A3，对CSI数据处理，建模，提取心室波动曲线、心房波动曲线和呼吸波动曲线；A4，提取情绪波动曲线，进行聚类分析；A5，通过深度学习识别内在情绪；A6，根据实测的准确率，返回步骤A4进行迭代分析。

Description

一种基于WiFi信号的情绪识别方法 技术领域
本发明属于人工智能技术领域，特别涉及一种基于WiFi信号的情绪识别方法。
背景技术
随着人工智能与机器人的快速发展，人机交互中对用户情绪识别的需求越来越强烈，大多数的情绪识别通过摄像头装置使用计算机视觉算法识别人脸情绪，但是这种方案存在着很大的局限性，往往不能够识别出用户的内在情绪。比如，通过计算机视觉检测到人脸呈现出微笑状态，判断这个人是高兴的，但是有些这样的“微笑状态”是负面的，可能是轻蔑，可能是无奈，可能是习惯性的平静。
最近几年有一些研究者提出了使用ECG(心电图)来识别人脸情绪，通过心律的变化去识别人的开心，高兴，生气，悲伤等情绪，并且具有比较高的精准率。然而，这种方案需要用户穿戴ECG设备，才能够检测到用户的内在情绪。
发明内容
本发明提出了一种基于WiFi信号的情绪识别方法，不需要用户穿戴任何设备，只需要安装在家庭、办公室等的普通WiFi设备。利用WiFi中CSI信号描述了信号在每条传输路径上的衰弱因子，比如信号散射，环境衰弱，距离衰减等信息，对微弱信号具有很高的敏感性。本发明通过WiFi中的CSI信号数据提取用户的偏强烈的正面情绪和负面情绪特征数据，识别出用户的内在真实情绪。
一种基于WiFi信号的情绪识别方法，该方法包括以下步骤：
A1，收集在不同情绪下，WiFi路由器发出的无线信号；
A2，从无线信号中提取出CSI信号；
A3，对CSI数据处理，建模，提取心室波动曲线、心房波动曲线和呼吸波动曲线；
A4，提取情绪波动曲线，进行聚类分析；
A5，通过深度学习识别内在情绪；
A6，根据实测的准确率，返回步骤A4进行迭代分析；
其中，对于步骤A3，对于CSI数据处理包括：
B1，数据预处理，包括：
针对由情绪引起的心率变动和呼吸变动是微小变动，使用带通滤波过滤高频数据；
针对不同WiFi设备发送功率对CSI数据的影响，对CSI数据进行归一化处理；
B2，去除环境噪声，包括：
多径抑制，即，将CSI从频域变换到时域(IFFT)，再通过移除与心率，呼吸不相关的长时延的信号成分，以减轻环境变化和人体移动引起的多径变化；
小波滤波，即，利用离散小波变换去除高频噪声；
B3，去除个体差异，即在收集的数据中排除与情绪无关的影响因素，包括性别、年龄、当前的活动量；
又有，步骤A4包括：
提取情绪波动曲线，即，通过对CSI频域数据特征分析，提取和心率，呼吸相关的信道数据，将这些周期性数据提取出来；并与实验室记录的情绪发生时间，持续时间匹配与对比，在统计上使用F检验，找出波动明显的数据；
接着进行情绪识别，即对心房波动曲线、心室波动曲线、呼吸曲线，分析IPI(Inter-Peak-Interval,波峰间隔)，SDIPI(即波峰间隔的标准差)；
并且分析不同情绪的功率谱表现，对这些数据做聚类分析，再分析与调整相关特征数据；
将所述相关特征数据转化成一个向量，作为输入，使用深度学习算法进行分类识别，并记录分类结果，与原始数据分析比较，分析精准率，召回率数据，对模型进行迭代。
本发明涉及到通信技术中WiFi中的CSI(Channel State Information，信道状态信息)信号关于不同情绪的特征提取，识别出用户真实情绪，相较于计算机视觉的情绪识别具有更高的准确性，可以应用于智能家居中的自然语言交互， 提高人机交互的情感特性，还可以应用于一些撒谎检测等等。
本发明通过室内安装的Wi-Fi路由器设备，通过解析WiFi中的CSI信号识别用户的偏强烈的情绪，在智能家居生活中，可以改善人机交互的情感能力，真实识别出用户的内在意图，另外可以主动式缓和用户当前的糟糕情绪，避免因为情绪不良引起的各类疾病，比如胃病，心脏病等等。
本发明的优势在于利用普通的WiFi设备比较准确地识别用户内在情绪。CSI信号具有比较高的多径分辨能力，尤其是对非视距范围内信号的微小变化具有很高的捕获能力，CSI信号对情绪识别也具有较高的灵敏度和感知广度。用户情绪与心率变化有关，而用户情绪在心率和呼吸上往往是一个上升，到达封顶，再衰减的过程，一个倒U的曲线，与心率，呼吸这样周期性的波动不一样；而不同情绪之间有很不一样的特征，通过特征指标会对这些数据进行聚类分析和有监督的分类识别。
与现有的技术相比，本发明的技术方案具有以下效果：
1)准确性：本发明通过CSI信号数据甄别用户的“心理活动”特征数据，将用户心跳加速，呼吸加速，两者持续时间长度等数据解析出来，识别用户内在偏强烈的情绪，比通过计算机视觉识别出的情绪更加准确；
2)灵敏性：通过对数据特征的提取，会转化出心房波动曲线，心室波动曲线，情绪波动曲线，IBI(inter-beat-interval,心搏间期)，SDIBI(心搏间期标准差)等数据信息，细致地分析出用户不同特征的内在情绪；
3)便捷性：用户不需要通过穿戴设备，比如ECG设备等来识别用户内在情绪，
4)高效性：如果用户在Wi-Fi设备识别情绪系统周边，能够一天24小时实时识别用户当前情绪，另一方面，比通过摄像头识别用户内在情绪的准确率更高，而且可以保护用户的肖像权等隐私。
附图说明
通过参考附图阅读下文的详细描述，本发明示例性实施方式的上述以及其他目的、特征和优点将变得易于理解。在附图中，以示例性而非限制性的方式示出了本发明的若干实施方式，其中：
图1是本发明实施例中涉及的WiFi情绪识别装置。
图2是本发明涉及到的情绪分类图。
图3是本发明实施例中基于WiFi信号的情绪识别方法流程图。
图4本发明实施例中情绪识别流程示意图。
具体实施方式
基于802.11n协议的WiFi技术采用MIMO-OFDM系统，利用华盛顿大学的Daniel Halperin提供的工具软件，可以获取一个无线通信信道中的30个子载波。只需要一个Wi-Fi路由器，一个远程服务器用于处理情绪识别，就可以在用户的手机，电脑等设备上查看自己的情绪识别。也可以直接将情绪识别工程运行在自己的手机上，PC上，并在类似的终端设备中实时查看识别结果。Wi-Fi情绪识别装置见图1。
基于802.11n协议的Wi-Fi技术采用MIMO-OFDM系统(MIMO即Multiple-Input Multiple-Output,多输入多输出；OFDM即Orthogonal Frequency Division Multiplexing，正交频分复用技术)，利用华盛顿大学的Daniel Halperin提供的工具软件，可以获取一个无线通信信道中的30个子载波；最后可以得到规范化的CSI矩阵H——n*m*30的复数矩阵，其中n表示发射天线个数，m表示接受天线个数，30是子载波信息个数。一个2*3*30的信道信息的模矩阵如下所示：
Figure PCTCN2018110206-appb-000001
本发明中对于情绪分类是基于罗素(Russell，1980)提出情绪分类的环状模式，将情绪划分为两个维度：愉快度和强度。其中愉快度分为愉快和不愉快，强度分为中等强度和高等强度，不考虑低强度的情绪识别，在此通过分类算法识别下图的八种情绪：惊奇，高兴；满足，疲乏；颓废，沮丧；生气，惊恐。
情绪数据采集实验设计包括：
实验者需求：年龄在19-77岁，20个女性，20个男性，年龄呈现均匀分布；
实验环境：两个一样布置黑暗的房间，两个一样布置明亮的房间，每个房间相同的WiFi设备，和信号接受处理器；
实验内容：每个实验对象，可以借助一些小道具在每个房间里面展现不同情绪，展现内容见下表：
情绪 惊奇 高兴 满足 疲乏 颓废 沮丧 生气 惊恐
站立                
坐着                
实验流程：
1)实验者先登记个人信息，情绪平稳，再带着提前设计好的情绪展现小卡片进房间；
2)按照一定顺序展现情绪，有工作人员会给出不同情绪Emoji和文字提示，当用户进入状态之后，提示工作人员，同时使用EGC设备测量实验者的心率数据；情绪维持时间是2分钟，再顺其自然进入平和状态；
3)按照同样的方式展示下一个情绪；
4)站立的情绪展现完成之后，休息10～15分钟，实验者坐在沙发上，按照同样的方式展现情绪；
5)第二天进入实验者在第二个房间进行实验，整个实验需要持续四天；
基于WIFi中CSI信号数据去识别情绪，设计的硬件设备简单，难点在于获取和情绪波动相关的心率波动曲线(心房，心室引起的多径效应)和呼吸波动曲线(胸腔引起的多径效应)，在频域数据上提取愉快，强烈特征相关的数据，通过深度学习算法来实现分类。模型训练的过程主要包括以下步骤：1.接受室内用户在不同情绪下WiFi路由器发出信号，收集WiFi上传的数据；2.从无线信号中提取出CSI信号；3.对CSI数据处理，包括数据预处理，去噪音等等，再关于心率周期性波动，呼吸周期性波动做数据特征。
其中的情绪识别建模与修正方法包括步骤：
1.数据预处理
由情绪引起的心率变动，呼吸变动是更加微小的变动，使用带通滤波过滤高频数据，使用2阶butterworth filter；另一方面由于不同Wi-Fi设备发送功率对CSI数据有一定影响，需要对数据进行归一化处理；
2.环境噪声去除
多径抑制：将CSI从频域变换到时域(IFFT)，再通过移除与心率，呼吸不相关的长时延的信号成分，以减轻环境变化和人体移动引起的多径变化；
小波滤波：利用离散小波变换去除高频噪声，其中threshholding部分采用动态阈值，4层Symlet小波；
3.个体差异去除
收集的数据会排除一些和情绪无关的影响因素，个体上的差异数据，比如性别，年龄，当前的活动量等等；
4.情绪识别波动曲线
通过对CSI频域数据特征分析，提取和心率，呼吸相关的信道数据，将这些周期性数据提取出来；并与实验室记录的情绪发生时间，持续时间匹配与对比，在统计上使用F检验，找出波动明显的数据；
5.情绪识别
再对这些心房波动曲线，心室波动曲线，呼吸曲线，分析IPI(Inter-Peak-Interval,波峰间隔)，SDIPI，即波峰间隔的标准差，时间窗是2分钟；并且分析不同情绪的功率谱表现，对这些数据做聚类分析，再分析与调整相关特征数据。再将这些特征数据转化成一个向量，作为输入，使用深度学习算法进行分类识别，并记录分类结果，与原始数据分析比较，分析精准率，召回率等数据，对模型进行迭代。
值得说明的是，虽然前述内容已经参考若干具体实施方式描述了本发明创造的精神和原理，但是应该理解，本发明并不限于所公开的具体实施方式，对各方面的划分也不意味着这些方面中的特征不能组合，这种划分仅是为了表述的方便。本发明旨在涵盖所附权利要求的精神和范围内所包括的各种修改和等同布置。

Claims (1)

  1. 一种基于WiFi信号的情绪识别方法，其特征在于，该方法包括以下步骤：
    A1，收集在不同情绪下，WiFi路由器发出的无线信号；
    A2，从无线信号中提取出CSI信号；
    A3，对CSI数据处理，建模，提取心室波动曲线、心房波动曲线和呼吸波动曲线；
    A4，提取情绪波动曲线，进行聚类分析；
    A5，通过深度学习识别内在情绪；
    A6，根据实测的准确率，返回步骤A4进行迭代分析；
    其中，对于步骤A3，对于CSI数据处理包括：
    B1，数据预处理，包括：
    针对由情绪引起的心率变动和呼吸变动是微小变动，使用带通滤波过滤高频数据；
    针对不同WiFi设备发送功率对CSI数据的影响，对CSI数据进行归一化处理；
    B2，去除环境噪声，包括：
    多径抑制，即，将CSI从频域变换到时域(IFFT)，再通过移除与心率，呼吸不相关的长时延的信号成分，以减轻环境变化和人体移动引起的多径变化；小波滤波，即，利用离散小波变换去除高频噪声；
    B3，去除个体差异，即在收集的数据中排除与情绪无关的影响因素，包括性别、年龄、当前的活动量；
    步骤A4包括：
    提取情绪波动曲线，即，通过对CSI频域数据特征分析，提取和心率，呼吸相关的信道数据，将这些周期性数据提取出来；并与实验室记录的情绪发生时间，持续时间匹配与对比，在统计上使用F检验，找出波动明显的数据；
    接着进行情绪识别，即对心房波动曲线、心室波动曲线、呼吸曲线，分析IPI(Inter-Peak-Interval,波峰间隔)，SDIPI(即波峰间隔的标准差)；
    并且分析不同情绪的功率谱表现，对这些数据做聚类分析，再分析与调整相关特征数据；
    将所述相关特征数据转化成一个向量，作为输入，使用深度学习算法进行分类识别，并记录分类结果，与原始数据分析比较，分析精准率，召回率数据，对模型进行迭代。
PCT/CN2018/110206 2017-10-23 2018-10-15 一种基于WiFi信号的情绪识别方法 WO2019080734A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710994059.0 2017-10-23
CN201710994059.0A CN107822645B (zh) 2017-10-23 2017-10-23 一种基于WiFi信号的情绪识别方法

Publications (1)

Publication Number Publication Date
WO2019080734A1 true WO2019080734A1 (zh) 2019-05-02

Family

ID=61648885

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/110206 WO2019080734A1 (zh) 2017-10-23 2018-10-15 一种基于WiFi信号的情绪识别方法

Country Status (2)

Country Link
CN (1) CN107822645B (zh)
WO (1) WO2019080734A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113855018A (zh) * 2021-08-09 2021-12-31 燕山大学 一种基于连续波声音信号情绪识别的方法
CN116074835A (zh) * 2023-01-16 2023-05-05 上海物骐微电子有限公司 基于WiFi的手势识别方法及装置、电子设备、存储介质

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107822645B (zh) * 2017-10-23 2020-01-17 上海百芝龙网络科技有限公司 一种基于WiFi信号的情绪识别方法
CN108683467A (zh) * 2018-05-22 2018-10-19 深圳市普威技术有限公司 信号检测方法、通信设备、采集设备和信号检测系统
CN110141259B (zh) * 2019-06-04 2021-09-28 清华大学 一种基于无线通信测量认知负荷与心理压力的方法及装置
CN113116294A (zh) * 2019-12-30 2021-07-16 上海际链网络科技有限公司 人员身体状况的监测方法及装置
CN114694244A (zh) * 2020-12-30 2022-07-01 华为技术有限公司 颈动数据采集与处理方法及装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105232022A (zh) * 2015-09-17 2016-01-13 太原理工大学 基于WiFi中CSI信号强度的非侵入式呼吸心跳检测实现方法
CN106066995A (zh) * 2016-05-25 2016-11-02 西安交通大学 一种无线非绑定人体行为检测算法
CN106175767A (zh) * 2016-07-01 2016-12-07 华中科技大学 一种非接触式的多人呼吸参数实时检测方法及系统
CN106725488A (zh) * 2016-12-27 2017-05-31 深圳大学 一种无线场强呼吸检测方法、装置及呼吸检测仪
WO2017156492A1 (en) * 2016-03-11 2017-09-14 Origin Wireless, Inc. Methods, apparatus, servers, and systems for vital signs detection and monitoring
CN107822645A (zh) * 2017-10-23 2018-03-23 上海百芝龙网络科技有限公司 一种基于WiFi信号的情绪识别方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105232022A (zh) * 2015-09-17 2016-01-13 太原理工大学 基于WiFi中CSI信号强度的非侵入式呼吸心跳检测实现方法
WO2017156492A1 (en) * 2016-03-11 2017-09-14 Origin Wireless, Inc. Methods, apparatus, servers, and systems for vital signs detection and monitoring
CN106066995A (zh) * 2016-05-25 2016-11-02 西安交通大学 一种无线非绑定人体行为检测算法
CN106175767A (zh) * 2016-07-01 2016-12-07 华中科技大学 一种非接触式的多人呼吸参数实时检测方法及系统
CN106725488A (zh) * 2016-12-27 2017-05-31 深圳大学 一种无线场强呼吸检测方法、装置及呼吸检测仪
CN107822645A (zh) * 2017-10-23 2018-03-23 上海百芝龙网络科技有限公司 一种基于WiFi信号的情绪识别方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZHAO, MINGMIN ET AL.: "Emotion Recognition using Wireless Signals", MOBICOM ' 16 PROCEEDINGS OF THE 22ND ANNUAL INTERNATIONAL CONFERENCE ON MOBILE COMPUTING AND NETWORKING, 7 October 2016 (2016-10-07), XP058279593 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113855018A (zh) * 2021-08-09 2021-12-31 燕山大学 一种基于连续波声音信号情绪识别的方法
CN116074835A (zh) * 2023-01-16 2023-05-05 上海物骐微电子有限公司 基于WiFi的手势识别方法及装置、电子设备、存储介质

Also Published As

Publication number Publication date
CN107822645A (zh) 2018-03-23
CN107822645B (zh) 2020-01-17

Similar Documents

Publication Publication Date Title
WO2019080734A1 (zh) 一种基于WiFi信号的情绪识别方法
CN108553108B (zh) 一种基于Wi-Fi中CSI信号的人体动作与呼吸的检测方法和系统
CN111329474B (zh) 基于深度学习的脑电身份识别方法、系统及信息更新方法
Komeili et al. Feature selection for nonstationary data: Application to human recognition using medical biometrics
Pan et al. Emotion recognition based on EEG using generative adversarial nets and convolutional neural network
CN101828921A (zh) 基于视觉诱发脑电vep的身份识别方法
EP3449409B1 (en) Biometric method and device for identifying a person through an electrocardiogram (ecg) waveform
CN111387975B (zh) 基于机器学习的脑电信号识别方法
CN114947852B (zh) 一种多模态情感识别方法、装置、设备及存储介质
CN108256579A (zh) 一种基于先验知识的多模态民族认同感量化测量方法
CN111000556A (zh) 一种基于深度模糊森林的情绪识别方法
CN113723557A (zh) 一种基于多频带时空卷积网络的抑郁脑电分类系统
Putra et al. EEG-based emotion classification using wavelet decomposition and K-nearest neighbor
Setiawan et al. A framework for real time emotion recognition based on human ans using pervasive device
CN114129163B (zh) 基于脑电信号的多视图深度学习的情绪分析方法和系统
KR101760441B1 (ko) 심전도 정점 주기를 이용한 개인 식별 장치 및 그 방법
Cai et al. Combination of EOG and EEG for emotion recognition over different window sizes
Aydemir Odor and Subject Identification Using Electroencephalography Reaction to Olfactory.
Kehri et al. A facial EMG data analysis for emotion classification based on spectral kurtogram and CNN
Wijayanto et al. Biometric identification based on EEG signal with photo stimuli using Hjorth descriptor
Tang et al. Eye movement prediction based on adaptive BP neural network
Kim et al. A study on user recognition using 2D ECG image based on ensemble networks for intelligent vehicles
CN113057654B (zh) 基于频率耦合神经网络模型的记忆负荷检测提取系统及方法
Markova et al. Detection of negative emotions and high-arousal negative-valence states on the move
Nawas et al. K-NN classification of brain dominance

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: 18871578

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18871578

Country of ref document: EP

Kind code of ref document: A1