WO2022108057A1 - Appareil multi-modal et procédé de classification d'émotions - Google Patents

Appareil multi-modal et procédé de classification d'émotions Download PDF

Info

Publication number
WO2022108057A1
WO2022108057A1 PCT/KR2021/011339 KR2021011339W WO2022108057A1 WO 2022108057 A1 WO2022108057 A1 WO 2022108057A1 KR 2021011339 W KR2021011339 W KR 2021011339W WO 2022108057 A1 WO2022108057 A1 WO 2022108057A1
Authority
WO
WIPO (PCT)
Prior art keywords
emotion classification
emotion
classification
real image
modal
Prior art date
Application number
PCT/KR2021/011339
Other languages
English (en)
Korean (ko)
Inventor
노희열
이선행
이유중
장동영
윤찬녕
신영하
Original Assignee
오로라월드 주식회사
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 오로라월드 주식회사 filed Critical 오로라월드 주식회사
Publication of WO2022108057A1 publication Critical patent/WO2022108057A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/64—Three-dimensional objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/172—Classification, e.g. identification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/174—Facial expression recognition
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
    • G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
    • G10L25/63—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination for estimating an emotional state

Definitions

  • the present invention relates to a multi-modal-based emotion classification apparatus and method.
  • the present disclosure is derived from research conducted as part of the WC300 R&D project of the Korea Institute of Industrial Technology Promotion. [Project unique number: 1425140847, task name: Development of interactive smart toy and expandable content service platform through emotion recognition/interactive technology]
  • an apparatus and method for verifying emotion classification predicted based on an image using emotion classification measured based on voice, and updating an emotion classification model based on verification accuracy There is a purpose.
  • the expression information of an object included in the real image is recognized from the real image, and expression information is extracted, and clustering-based emotion Predict the emotional classification of the object included in the real image using a classification model, extract temperature information of the object included in the thermal image from the thermal image, and use the emotion classification model to predict the object included in the thermal image a classification prediction unit for predicting the classification of emotions; a classification measurement unit for measuring an emotion classification represented by the voice data from the voice data using a band-pass filter for filtering a frequency band according to the emotion classification of an object; and verifying each of the emotion classification predicted based on the real image and the emotion classification predicted based on the thermal image using the emotion classification measured based on the voice data, and the emotion classification model based on the verification result
  • It provides a multi-modal-based emotion classification apparatus, characterized in that it includes an updater for updating.
  • the expression information of an object included in the real image is recognized from the real image, and expression information is extracted, and clustering-based emotion predicting the emotional classification of the object included in the real image from the classification model; extracting temperature information of the object included in the thermal image from the thermal image and predicting the emotional classification of the object included in the thermal image from the emotion classification model; measuring an emotion classification represented by the voice data from the voice data using a band-pass filter that filters a frequency band according to the emotion classification of an object;
  • Each of the emotion classification predicted based on the real image and the emotion classification predicted based on the thermal image is verified using the emotion classification measured based on the voice data, and the emotion classification model is updated based on the verification result. It provides a multi-modal-based emotion classification method comprising the process of:
  • the update unit is based on the verification accuracy or verification reliability of the emotion classification predicted for the first value and the emotion classification predicted for the second value to update the emotion classification model by updating a weight corresponding to each value.
  • the updater changes the weight of the emotion classification model
  • the classification prediction unit uses the changed emotion classification model to provide a multi-modal-based emotion classification apparatus, characterized in that the emotion classification is re-predicted based on the expression information and the temperature information.
  • emotion classification predicted based on a normal image and a thermal image is verified using emotion classification measured based on voice, and emotion classification is based on verification accuracy
  • emotion classification is based on verification accuracy
  • emotion by calculating the verification accuracy based on the emotion classification calculated for each modality, by changing the weight corresponding to each modality of the emotion classification model employed by the emotion classification apparatus and method, emotion It has the effect of improving the accuracy and/or reliability of classification.
  • FIG. 1 is a block diagram illustrating an emotion classification apparatus according to an embodiment of the present disclosure.
  • FIG. 2 is a conceptual diagram illustrating a method of predicting emotion classification from an image according to an embodiment of the present disclosure.
  • FIG. 3 is a conceptual diagram illustrating a method of measuring emotion classification from voice data according to an embodiment of the present disclosure.
  • FIG. 4 is a flowchart illustrating a method for classifying emotions according to an embodiment of the present disclosure.
  • FIG 5 is an exemplary diagram of an emotion classification result calculated based on the emotion classification apparatus according to an embodiment of the present disclosure.
  • An image in the present disclosure includes both a still image and a video including an object, for example, a human face.
  • FIG. 1 is a block diagram illustrating an emotion classification apparatus according to an embodiment of the present disclosure.
  • the emotion classification apparatus 100 includes all or part of a classification prediction unit 102 , a classification measurement unit 104 and an update unit 106 .
  • the emotion classification apparatus 100 shown in Fig. 1 is according to an embodiment of the present disclosure, and not all components shown in Fig. 1 are essential components, and some components may be added, changed, or deleted in other embodiments.
  • the emotion classification apparatus may further include a display unit (not shown) for displaying the verified and predicted emotion classification results.
  • the noise removing device is a software module that performs the functions of each component 102 to 106 Alternatively, it may be implemented as a processor.
  • the classification prediction unit 102 recognizes the expression of an object included in the real image from the real image and extracts expression information Then, the temperature information of the object included in the thermal image is extracted from the thermal image, and the emotion classification of each object is predicted from the clustering-based emotion classification model. Specifically, the classification prediction unit 102 predicts the emotion classification for each of the first value obtained by digitizing the expression information extracted from the real image and the second value that is the temperature information extracted from the thermal image using the emotion classification model.
  • the emotion classification prediction of the emotion classification model may be performed by predicting a probability corresponding to each emotion classification with respect to one or more emotion classifications.
  • emotional classifications may be happy, sad, scary, angry, surprised, etc., and may further include neutral, which is an unclassified state that is not included anywhere. does not limit For example, in relation to the classification of emotions, such as touching, hopeful, playful, satisfied, confused, hateful, envious, etc., the emotion classification model is usually Any sentiment classification that can be used for
  • the classification prediction unit 102 extracts the position of the face of the object from the real image, and based on the extracted face position, information on face landmarks (eg, eyes, nose, mouth, cheeks, forehead, etc.) (eg: location, area, etc.) to obtain a first value from a pre-learned expression value extraction model.
  • the facial landmark information may be obtained by converting a real image to gray scale and extracting a boundary line, that is, an edge of the image.
  • the first value converts the extracted edge value into a two-dimensional matrix, transforms the converted two-dimensional matrix into a three-dimensional matrix using the camera focus value, and converts the transformed three-dimensional matrix into a pre-learned expression value extraction model It can be obtained by entering
  • the classification prediction unit 102 sets one or more regions of interest (ROI) based on the face position of the object extracted from the real image, and each region of interest The temperature for is extracted as the second value.
  • the second value refers to the position of the face recognized by the visible light camera and gives weight to the position of the face to obtain a thermal image corresponding to the face, and then, from the thermal image of the face, the region of interest corresponding to the facial landmark It can be obtained by extracting the temperature.
  • the classification measurement unit 104 measures emotion classification from voice data using a band-pass filter that filters a frequency band according to the emotional state.
  • the band filter may be generated based on a frequency average value according to emotion classification.
  • the band filter converts the same speech data as the first graph in which the speech data is Fast Fourier Transform (FFT) based on amplitude and frequency, based on time and amplitude, and performs a short-time Fourier transform (STFT, Short-Time Fourier Transform) may be generated by analyzing common characteristics (eg, graph reformation in the same area) between the second graphs.
  • FFT Fast Fourier Transform
  • STFT Short-Time Fourier Transform
  • the update unit 106 verifies the emotion classification predicted based on the image including the real image and the thermal image using the emotion classification measured based on the voice data, and updates the emotion classification model based on the verification accuracy. Specifically, this update verifies the emotion classification predicted by the emotion prediction unit 102 using the emotion classification measured by the emotion measurement unit 104, and each modality of the emotion classification model based on the verification accuracy This may be performed by updating a weight corresponding to (modality). As another example, the emotion prediction unit 102 corresponds to each value of the emotion classification model based on the verification accuracy of each of the emotion classification predicted for the first value and the emotion classification predicted for the second value This can be done by updating the weights.
  • the classification prediction unit 102 re-predicts the emotion classification, and the update unit 106 verifies the re-predicted emotion classification to say the It is desirable to update the emotion classification model.
  • FIG. 2 is a conceptual diagram illustrating a method of predicting emotion classification from an image according to an embodiment of the present disclosure.
  • the emotion classification apparatus When the emotion classification apparatus obtains a real image from a camera or a visible light sensor, it recognizes the position of the face from the real image and extracts features from the face.
  • the position of the face is determined by processing signals and/or data using, for example, all or part of a real image normalization, binarization, outline extraction, and recognition classifier (eg, Haar, HOG, etc.). can be recognized
  • a real image normalization, binarization, outline extraction, and recognition classifier eg, Haar, HOG, etc.
  • the original real image is converted to gray scale or the real image processed with signal and/or data is converted to gray scale to track the position of the face, and boundary lines are extracted from the converted image to extract the eyes, nose, mouth, and forehead. , clown, and other facial landmarks are extracted.
  • the emotion classification apparatus converts the facial landmark information into a two-dimensional matrix, and converts the converted two-dimensional matrix into a three-dimensional matrix by using a focus value of a camera or a visible light sensor that obtained a real image.
  • the emotion classification device extracts a value for a facial expression from the transformed three-dimensional matrix using a pre-learned expression value extraction model, and uses the emotion classification model to determine a probability value belonging to each emotion classification that the emotion classification model can classify. predict
  • the emotion classification device acquires a thermal image from a thermal imaging camera, etc. Temperature information can be obtained for areas such as eyes and mouth.
  • Temperature information can be obtained for areas such as eyes and mouth.
  • the emotion classification device recognizes the region where the face is located from the entire thermal image by giving weight to the position of the image with reference to position information of the face recognized by a camera or a visible light sensor, for example. .
  • the emotion classification device sets a region of interest (ROI) using the region where the recognized face is located, extracts the temperature of the regions corresponding to the eyes, nose, mouth, cheeks, forehead, etc., and uses the emotion classification model to Predict the probability value belonging to each emotion classification that the emotion classification model can classify.
  • ROI region of interest
  • FIG. 3 is a conceptual diagram illustrating a method of measuring emotion classification from voice data according to an embodiment of the present disclosure.
  • the emotion classification apparatus obtains voice information corresponding to various emotion classifications from a microphone or the like, and processes the voice information to generate a band-pass filter.
  • the emotion classification apparatus obtains voice data by converting voice information into digital information to maintain the characteristics of the frequency domain. Thereafter, noise suppression is performed in the frequency domain of the voice data using a noise removal filter (eg, a Gaussian filter, etc.), and the pitch of the voice data or the noise-removed voice data , word clause, and speed of speech are extracted and quantified.
  • a noise removal filter eg, a Gaussian filter, etc.
  • FFT Fast Fourier Transform
  • STFT Short Time Fourier Transform
  • the emotion classification apparatus Fourier transforms speech data to be used for verification, applies a bandpass filter to the transformed speech data, and measures a probability value of each emotion classification that the bandpass filter can classify.
  • FIG. 4 is a flowchart illustrating a method for classifying emotions according to an embodiment of the present disclosure.
  • the emotion classification apparatus obtains a real image and a thermal image corresponding to the same emotion classification, and predicts the emotion classification using the obtained real image (S400).
  • An emotion classification model is used for prediction of emotion classification.
  • the emotion classification apparatus acquires voice data corresponding to the same emotion classification as the real image and thermal image in parallel with step S400 or before/after step S400, and applies a band filter to the obtained voice data to classify emotions is measured (S402). Such measurement may be performed by calculating a probability value corresponding to each emotion classification.
  • the emotion classification apparatus verifies the emotion classification predicted based on images, such as real images and thermal images, and uses emotion classification measured based on voice, such as voice data, for verification ( S404 ).
  • the verification is performed by determining whether the predicted emotion classification and the measured emotion classification match within an error range (S406).
  • Such verification is preferably performed in consideration of preset verification conditions (eg, reliability, accuracy, etc.).
  • step S406 If it is determined in step S406 that the predicted emotion classification and the measured emotion classification match within the error range, the predicted emotion classification is calculated as an emotion classification result and the procedure is terminated.
  • the emotion classification apparatus changes a weight corresponding to each modality of the emotion classification model (S408). Thereafter, the emotion classification is re-predicted using the real image and the thermal image obtained in step S400 using the changed emotion classification model.
  • FIG 5 is an exemplary diagram of an emotion classification result calculated based on the emotion classification apparatus according to an embodiment of the present disclosure.
  • the emotion classification apparatus may calculate an emotion classification result of the emotion classification model as shown in FIG. 5 .
  • the calculation of the emotion classification result is preferably performed in real time. For example, as shown in FIG. 5 , real images, thermal images, and voice data are obtained from an object in real time, and expression values and temperatures are extracted from the real images and thermal images, respectively, and emotion classification is predicted from the emotion classification model. Thereafter, the emotion classification apparatus measures the emotion classification corresponding to the speech data by Fourier transforming the speech data and passing it through a band filter. The predicted emotion classification is verified based on the measured emotion classification, and when the validation result is validated, a percentage corresponding to each emotion classification is calculated as a data value or a graph as shown in FIG. 5 . This calculation can be performed only for some emotion classification results that have been validated. In addition, if the emotion classification corresponding to 'unclassified' is greater than or equal to a preset percentage, this calculation may not be performed because the verification result is determined to be invalid.
  • Various implementations of the devices, units, processes, steps, etc., described herein may include digital electronic circuits, integrated circuits, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), computer hardware, firmware, software, and/or a combination thereof. These various implementations may include being implemented in one or more computer programs executable on a programmable system.
  • the programmable system includes at least one programmable processor (which may be a special purpose processor) coupled to receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device. or may be a general-purpose processor).
  • Computer programs also known as programs, software, software applications or code
  • the computer-readable recording medium includes all types of recording devices in which data readable by a computer system is stored. These computer-readable recording media are non-volatile or non-transitory, such as ROM, CD-ROM, magnetic tape, floppy disk, memory card, hard disk, magneto-optical disk, and storage device. It may further include a medium or a transitory medium such as a data transmission medium. In addition, the computer-readable recording medium may be distributed in network-connected computer systems, and computer-readable codes may be stored and executed in a distributed manner.
  • the computer includes a programmable processor, a data storage system (including volatile memory, non-volatile memory, or other types of storage systems or combinations thereof), and at least one communication interface.
  • the programmable computer may be one of a server, a network appliance, a set-top box, an embedded device, a computer expansion module, a personal computer, a laptop, a Personal Data Assistant (PDA), a cloud computing system, or a mobile device.
  • PDA Personal Data Assistant

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Data Mining & Analysis (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Signal Processing (AREA)
  • Computational Linguistics (AREA)
  • Psychiatry (AREA)
  • Acoustics & Sound (AREA)
  • Hospice & Palliative Care (AREA)
  • Child & Adolescent Psychology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Image Analysis (AREA)

Abstract

La présente divulgation concerne un appareil et un procédé de classification d'émotions. Un aspect de la présente invention concerne un appareil et un procédé, dans lequel des émotions classées sur la base d'images sont vérifiées à l'aide de classifications d'émotions déterminées sur la base du son et, sur la base de la précision de la vérification, un modèle de classification d'émotions est mis à jour.
PCT/KR2021/011339 2020-11-20 2021-08-25 Appareil multi-modal et procédé de classification d'émotions WO2022108057A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
KR1020200157046A KR102428916B1 (ko) 2020-11-20 2020-11-20 멀티-모달 기반의 감정 분류장치 및 방법
KR10-2020-0157046 2020-11-20

Publications (1)

Publication Number Publication Date
WO2022108057A1 true WO2022108057A1 (fr) 2022-05-27

Family

ID=81709314

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2021/011339 WO2022108057A1 (fr) 2020-11-20 2021-08-25 Appareil multi-modal et procédé de classification d'émotions

Country Status (2)

Country Link
KR (1) KR102428916B1 (fr)
WO (1) WO2022108057A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
ES2957419A1 (es) * 2022-06-03 2024-01-18 Neurologyca Science & Marketing Sl Sistema y método de detección en tiempo real de estados emocionales mediante visión artificial y escucha del lenguaje natural

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102695645B1 (ko) * 2023-10-11 2024-08-16 주식회사 수피야 근로자의 실시간 감정 분석 및 임무수행적합성 판별 플랫폼 서비스 제공 시스템

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20120092846A (ko) * 2011-02-14 2012-08-22 충남대학교산학협력단 멀티 모달리티 감정인식 시스템, 감정인식 방법 및 그 기록매체
WO2016170810A1 (fr) * 2015-04-23 2016-10-27 ソニー株式会社 Dispositif de traitement d'informations, procédé de commande, et programme
KR20170006355A (ko) * 2015-07-08 2017-01-18 주식회사 케이티 모션벡터 및 특징벡터 기반 위조 얼굴 검출 방법 및 장치
JP2018068618A (ja) * 2016-10-28 2018-05-10 株式会社東芝 感情推定装置、感情推定方法、感情推定プログラム、および感情カウントシステム
KR20190023749A (ko) * 2017-08-30 2019-03-08 (주)휴머노이드시스템 사용자 감성 인식 장치 및 방법

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102429822B1 (ko) 2016-02-12 2022-08-08 한국전자통신연구원 운전자의 감정상태 인식 장치 및 방법

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20120092846A (ko) * 2011-02-14 2012-08-22 충남대학교산학협력단 멀티 모달리티 감정인식 시스템, 감정인식 방법 및 그 기록매체
WO2016170810A1 (fr) * 2015-04-23 2016-10-27 ソニー株式会社 Dispositif de traitement d'informations, procédé de commande, et programme
KR20170006355A (ko) * 2015-07-08 2017-01-18 주식회사 케이티 모션벡터 및 특징벡터 기반 위조 얼굴 검출 방법 및 장치
JP2018068618A (ja) * 2016-10-28 2018-05-10 株式会社東芝 感情推定装置、感情推定方法、感情推定プログラム、および感情カウントシステム
KR20190023749A (ko) * 2017-08-30 2019-03-08 (주)휴머노이드시스템 사용자 감성 인식 장치 및 방법

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
ES2957419A1 (es) * 2022-06-03 2024-01-18 Neurologyca Science & Marketing Sl Sistema y método de detección en tiempo real de estados emocionales mediante visión artificial y escucha del lenguaje natural

Also Published As

Publication number Publication date
KR20220069677A (ko) 2022-05-27
KR102428916B1 (ko) 2022-08-03

Similar Documents

Publication Publication Date Title
JP6938784B2 (ja) オブジェクト識別の方法及びその、コンピュータ装置並びにコンピュータ装置可読記憶媒体
WO2022108057A1 (fr) Appareil multi-modal et procédé de classification d'émotions
CN108898086B (zh) 视频图像处理方法及装置、计算机可读介质和电子设备
CN106683680B (zh) 说话人识别方法及装置、计算机设备及计算机可读介质
CN112233698B (zh) 人物情绪识别方法、装置、终端设备及存储介质
CN112149615B (zh) 人脸活体检测方法、装置、介质及电子设备
CN111091536B (zh) 医学图像处理方法、装置、设备、介质以及内窥镜
CN108229375B (zh) 用于检测人脸图像的方法和装置
KR20190108804A (ko) 주파수 변화에 강인한 음향 이벤트 검출 방법 및 그 장치
CN108877787A (zh) 语音识别方法、装置、服务器及存储介质
CN111563480A (zh) 冲突行为检测方法、装置、计算机设备和存储介质
CN111950570B (zh) 目标图像提取方法、神经网络训练方法及装置
WO2020238321A1 (fr) Procédé et dispositif d'identification d'âge
CN115620384B (zh) 模型训练方法、眼底图像预测方法及装置
CN111126159A (zh) 用于实时跟踪行人的方法、装置、电子设备和介质
CA3164373A1 (fr) Procede et appareil de detection de toux automatique
Hassan et al. Intelligent sign language recognition using enhanced fourier descriptor: a case of Hausa sign language
CN114882576B (zh) 人脸识别方法、电子设备、计算机可读介质和程序产品
CN114944152A (zh) 车辆鸣笛音识别方法
CN115222966A (zh) 对抗数据生成方法、装置、计算机设备及存储介质
CN112070022A (zh) 人脸图像识别方法、装置、电子设备和计算机可读介质
CN112633496A (zh) 一种检测模型的处理方法及装置
WO2019199035A1 (fr) Système et procédé de suivi du regard
CN115662447B (zh) 一种基于多特征融合的测谎分析方法及装置
CN112115740A (zh) 用于处理图像的方法和装置

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: 21894830

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: 21894830

Country of ref document: EP

Kind code of ref document: A1