WO2021036664A1 - 基于微表情的客户满意度识别方法、装置、终端及介质 - Google Patents

基于微表情的客户满意度识别方法、装置、终端及介质 Download PDF

Info

Publication number
WO2021036664A1
WO2021036664A1 PCT/CN2020/105631 CN2020105631W WO2021036664A1 WO 2021036664 A1 WO2021036664 A1 WO 2021036664A1 CN 2020105631 W CN2020105631 W CN 2020105631W WO 2021036664 A1 WO2021036664 A1 WO 2021036664A1
Authority
WO
WIPO (PCT)
Prior art keywords
satisfaction
score
satisfaction score
customer
recognition model
Prior art date
Application number
PCT/CN2020/105631
Other languages
English (en)
French (fr)
Inventor
严月强
Original Assignee
深圳壹账通智能科技有限公司
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 深圳壹账通智能科技有限公司 filed Critical 深圳壹账通智能科技有限公司
Publication of WO2021036664A1 publication Critical patent/WO2021036664A1/zh

Links

Images

Classifications

    • 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
    • G06V40/176—Dynamic expression
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/01—Customer relationship services
    • 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
    • G06V40/20—Movements or behaviour, e.g. gesture 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

  • This application relates to the field of artificial intelligence technology, and in particular to a method, device, terminal and medium for identifying customer satisfaction based on micro-expression.
  • the first aspect of the present application provides a method for identifying customer satisfaction based on micro-expressions, the method including:
  • the expression features are input into the pre-trained expression satisfaction recognition model
  • the body features are input into the pre-trained body satisfaction recognition model
  • the speech rate features and pitch features are input into the pre-training.
  • the final satisfaction degree of the customer is calculated and output according to the first satisfaction score, the second satisfaction score, and the third satisfaction score.
  • the second aspect of the present application provides a device for identifying customer satisfaction based on micro-expressions, the device comprising:
  • the first acquisition module is used to acquire the customer's full-body image and audio data of a preset duration every preset collection period;
  • An extraction module for extracting expression features and body features of predetermined points from the whole body image, and extracting speech rate features and pitch features from the audio data;
  • the input module is used to simultaneously input the expression features into the pre-trained facial expression satisfaction recognition model, input the limb characteristics into the pre-trained limb satisfaction recognition model, and combine the speech rate characteristics and pitch
  • the features are input into the pre-trained speech satisfaction recognition model
  • the second acquisition module is configured to acquire the first satisfaction score output by the facial expression satisfaction recognition model, the second satisfaction score output by the limb satisfaction recognition model, and the third satisfaction output by the speech satisfaction recognition model Degree points
  • the calculation module is configured to calculate and output the final satisfaction degree of the customer according to the first satisfaction score, the second satisfaction score, and the third satisfaction score.
  • a third aspect of the present application provides a terminal, the terminal includes a processor, and the processor is configured to implement the following steps when executing computer-readable instructions stored in a memory:
  • the expression features are input into the pre-trained expression satisfaction recognition model
  • the body features are input into the pre-trained body satisfaction recognition model
  • the speech rate features and pitch features are input into the pre-training.
  • the final satisfaction degree of the customer is calculated and output according to the first satisfaction score, the second satisfaction score, and the third satisfaction score.
  • a fourth aspect of the present application provides a computer-readable storage medium having computer-readable instructions stored on the computer-readable storage medium, and when the computer-readable instructions are executed by a processor, the following steps are implemented:
  • the expression features are input into the pre-trained expression satisfaction recognition model
  • the body features are input into the pre-trained body satisfaction recognition model
  • the speech rate features and pitch features are input into the pre-training.
  • the final satisfaction degree of the customer is calculated and output according to the first satisfaction score, the second satisfaction score, and the third satisfaction score.
  • the micro-expression-based customer satisfaction recognition method, device, terminal, and medium described in this application can be applied in fields such as smart government affairs, thereby promoting the development of smart cities.
  • This application collects customer’s facial expression characteristics, body characteristics, speech rate characteristics, and tone characteristics during the service process, and then uses multiple satisfaction recognition models to perform analysis on the expression characteristics, body characteristics, speech rate characteristics, and tone characteristics. Identify, get different satisfaction scores, and finally calculate the final satisfaction based on the different satisfaction scores.
  • the customer’s satisfaction cannot be calculated when the customer is expressionless during the entire service process, and by adopting multiple features, comprehensive consideration of the customer’s all-round information is adopted.
  • the calculated satisfaction is more realistic, which improves the success rate of satisfaction collection and ensures the accuracy of satisfaction collection.
  • FIG. 1 is a flowchart of a method for identifying customer satisfaction based on micro-expressions provided in Embodiment 1 of the present application.
  • Fig. 2 is a structural diagram of a micro-expression-based customer satisfaction recognition device provided in the second embodiment of the present application.
  • FIG. 3 is a schematic structural diagram of a terminal provided in Embodiment 3 of the present application.
  • FIG. 1 is a flowchart of a method for identifying customer satisfaction based on micro-expressions provided in Embodiment 1 of the present application.
  • the micro-expression-based customer satisfaction recognition method can be applied to terminals.
  • the micro-expression-based customer satisfaction provided by the method of this application can be directly integrated on the terminal.
  • the function of degree recognition may be run in the terminal in the form of a software development kit (SKD).
  • the method for identifying customer satisfaction based on micro-expression specifically includes the following steps. According to different needs, the order of the steps in the flowchart can be changed, and some can be omitted.
  • S11 Acquire a full-body image of the client and audio data of a preset duration every preset collection period.
  • the collection period can be preset, for example, every 5 seconds or 10 seconds can be set as a collection period.
  • the pre-installed camera is controlled to obtain the customer's full-body image every preset period. While obtaining the customer's full-body image, it also collects a piece of audio data of the customer's preset duration.
  • the first collection period and the second collection period may be preset, the whole body image of the client is collected in the first collection period, and the audio data of the client is collected in the second collection period.
  • Facial expression is a form of body language that can complete fine information communication.
  • the human face has ten muscles, which can produce extremely rich expressions and accurately convey a variety of different mentalities and emotions.
  • the customer's affirmation and emotion can be determined through facial expressions. Emotions such as negation, pleasure and disappointment, satisfaction and dissatisfaction.
  • the key parts of satisfactory performance are the mouth, cheeks, eyebrows, and forehead, and the key parts of dissatisfied performance are mouth and brow.
  • the corners of the mouth, the face of the eyes, the eyebrows, the forehead, and the chin can be preset as characteristic points, that is, the corners of the mouth, the face of the eyes, the eyebrows, the forehead, and the chin are predetermined points.
  • the extracting the facial expression features of the predetermined points from the full-body image includes: detecting a human face from the full-body image according to a preset face detection algorithm; and extracting the facial expression features of the predetermined points in the human face .
  • the speed of speech and the height of the pitch can express different emotions. For example, when a person is angry, the speed of speech is higher and the pitch is higher; when a person is happy, the speed of speech is slower and the pitch is moderate; In the case of sadness, the speech rate is slower and the pitch is lower. Therefore, after acquiring the audio data of the client's preset duration, the speech rate and pitch in the audio data are extracted.
  • the facial expression satisfaction recognition model, body satisfaction recognition model, and voice satisfaction recognition model are all pre-trained satisfaction recognition models. After the facial expression features, body features, speech rate features, and tone features are obtained At the same time, the expression characteristics, body characteristics, speech rate characteristics and tone characteristics are respectively input into the expression satisfaction recognition model, the body satisfaction recognition model, and the speech satisfaction recognition model for satisfaction recognition.
  • the training process of the facial expression satisfaction recognition model includes:
  • test pass rate is greater than or equal to the preset pass rate threshold, end the training of the facial expression satisfaction recognition model; otherwise, when the test pass rate is less than the preset pass rate threshold, retrain the expression Satisfaction recognition model until the test pass rate is greater than or equal to the preset pass rate threshold.
  • the training process of the body satisfaction recognition model and the speech satisfaction recognition model is the same as the facial expression satisfaction recognition process, and will not be elaborated here.
  • different satisfaction scores are preset according to different expression characteristics, body characteristics, speech speed characteristics, and tone characteristics. For example, a happy expression corresponds to a satisfaction score of 5 points, and an angry expression corresponds to satisfaction. The degree is -5 points. For ease of presentation, the satisfaction score corresponding to the expressionlessness is recorded as 0 points. The satisfaction score corresponding to fast speech and high pitch is -5 points; the satisfaction score corresponding to slow speech and moderate pitch is 5 points.
  • the user's facial expression characteristics, body characteristics, speech rate characteristics, tone characteristics, and satisfaction scores can be used as new data to increase the number of data sets, and retrain based on the new data sets Expression satisfaction recognition model, body satisfaction recognition model, and speech satisfaction recognition model, so as to continuously improve the recognition rate of each satisfaction recognition model.
  • S14 Obtain a first satisfaction score output by the facial expression satisfaction recognition model, a second satisfaction score output by the limb satisfaction recognition model, and a third satisfaction score output by the voice satisfaction recognition model.
  • the first satisfaction score can be output through the expression satisfaction recognition model, and the first satisfaction score represents the value corresponding to the customer’s expression characteristics. Satisfaction situation.
  • the second satisfaction score can be output through the body satisfaction recognition model, and the second satisfaction score represents the satisfaction degree corresponding to the client's body characteristics happening.
  • the third satisfaction score can be output through the speech satisfaction recognition model, and the third satisfaction score represents the customer’s speech rate feature and tone The satisfaction level corresponding to the feature. Different satisfaction scores represent different satisfaction situations.
  • S15 Calculate and output the final satisfaction degree of the customer according to the first satisfaction score, the second satisfaction score, and the third satisfaction score.
  • the first satisfaction score, the second satisfaction score, and the third satisfaction score are added and averaged to obtain the final satisfaction of the customer.
  • the customer's final satisfaction is calculated based on the expression characteristics, body characteristics, speech rate characteristics, and tone characteristics. It effectively combines the customer's all-round information, and the satisfaction obtained has more reference significance. And when the customer has no expression, no voice interaction, or no body, the satisfaction can still be calculated.
  • the calculating the final satisfaction degree of the customer according to the first satisfaction score, the second satisfaction score, and the third satisfaction score includes:
  • the first final satisfaction degree, the second final satisfaction degree, and the third final satisfaction degree are added and averaged to obtain the final satisfaction degree.
  • the facial expression features can most intuitively express the customer’s emotions, and the physical features have a certain inertia, it can be preset that the first weight value corresponding to the facial features is the largest and the second weight value corresponding to the physical characteristics is the smallest. , The third weight value corresponding to the speech rate feature and the tone feature is centered. The sum of the first weight value, the second weight value, and the third weight value is 1.
  • the facial expression satisfaction recognition model can output 12 first satisfaction scores and 12 first satisfaction scores.
  • the 12 first satisfaction scores are added up and divided by 12 to get the first average satisfaction score.
  • the second average satisfaction score and the third average satisfaction score can be calculated.
  • the final satisfaction degree is calculated according to the first average satisfaction score and the first weight value, the second average satisfaction score and the second weight value, the third average satisfaction score and the third weight value.
  • the final satisfaction calculated using statistical methods represents an overall satisfaction in the service process.
  • the method further includes:
  • the customer service is alerted according to the preset warning mode.
  • the degree of satisfaction of the customer service process is divided into four levels, the first level: the customer is very satisfied; the second level: the customer is relatively satisfied; the third level: the customer is basically satisfied; the fourth level: the customer is very satisfied .
  • Different levels of satisfaction correspond to different satisfaction scores.
  • the satisfaction score corresponding to the first level is 15-20 points, the satisfaction score corresponding to the first level is 10-15 points; the third level corresponds to satisfaction
  • the degree score is 5-10 points, and the satisfaction score corresponding to the fourth level is 0-5 points.
  • Set a satisfaction threshold in advance for example, 5 points. When the final satisfaction is less than 5 points, it is determined that the customer is very dissatisfied, and the customer service is alerted according to the preset alerting method.
  • the preset alarm mode may be to display the alarm content on the display screen of the customer service; or send the alarm information by email or short message.
  • the final satisfaction is obtained, and when the final satisfaction is lower than the preset satisfaction threshold, the customer service is alerted, which helps to improve the customer's subsequent service quality.
  • the method further includes:
  • the number of records increases by 1;
  • the full-body image and voice fragments of the customer are obtained from the first collection period, and the customer’s satisfaction scores in the first collection period are identified according to multiple satisfaction recognition models.
  • the number of records When there is a satisfaction score When the value is less than the preset satisfaction score, the number of records will be increased by 1.
  • the subsequent collection period if the number of records increases, when the number of records is greater than the threshold of the number of records, it indicates that in the process of customer service, the customer has clearly shown unsatisfactory emotions and unsatisfactory emotions. The number of times is too high. At this time, the customer service needs to be alerted, so that the customer service can improve the quality of service, and prevent the customer's dissatisfaction from erupting out of control.
  • the method further includes:
  • the lowest first target satisfaction score is selected from the first satisfaction score, the lowest second target satisfaction score, and the third satisfaction score are selected from the second satisfaction score
  • the lowest third goal satisfaction score is selected from the scores
  • the video stream of the entire service of the customer service is also captured by the camera device.
  • the satisfaction score output by the satisfaction recognition model will also be very low.
  • the video clip corresponding to this moment is extracted and sent to the customer for viewing and analysis, which is convenient for improving behaviors such as improper speech during follow-up services. Improve service quality.
  • the micro-expression-based customer satisfaction identification method described in this application can be applied in fields such as smart government affairs, thereby promoting the development of smart cities.
  • This application collects customer’s facial expression characteristics, body characteristics, speech rate characteristics, and tone characteristics during the service process, and then uses multiple satisfaction recognition models to perform analysis on the expression characteristics, body characteristics, speech rate characteristics, and tone characteristics. Identify, get different satisfaction scores, and finally calculate the final satisfaction based on the different satisfaction scores.
  • the customer’s satisfaction cannot be calculated when the customer is expressionless during the entire service process, and by adopting multiple features, comprehensive consideration of the customer’s all-round information is adopted.
  • the calculated satisfaction is more realistic, which improves the success rate of satisfaction collection and ensures the accuracy of satisfaction collection.
  • Fig. 2 is a structural diagram of a micro-expression-based customer satisfaction recognition device provided in the second embodiment of the present application.
  • the micro-expression-based customer satisfaction recognition device 20 may include a plurality of functional modules composed of computer-readable instructions.
  • the computer-readable instructions of each program segment in the micro-expression-based customer satisfaction recognition device 20 may be stored in the memory of the terminal and executed by the at least one processor for execution (see Figure 1 for details) The function of customer satisfaction recognition based on micro-expressions.
  • the micro-expression-based customer satisfaction recognition device 20 can be divided into multiple functional modules according to the functions it performs.
  • the functional modules may include: a first acquisition module 201, an extraction module 202, an input module 203, a training module 204, a second acquisition module 205, a calculation module 206, an alarm module 207, and a screening module 208.
  • the module referred to in this application refers to a series of computer-readable instructions that can be executed by at least one processor and can complete fixed functions, and are stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.
  • the first acquisition module 201 is configured to acquire a full-body image of the client and audio data of a preset duration every preset collection period.
  • the collection period can be preset, for example, every 5 seconds or 10 seconds can be set as a collection period.
  • the pre-installed camera is controlled to obtain the customer's full-body image every preset period. While obtaining the customer's full-body image, it also collects a piece of audio data of the customer's preset duration.
  • the first collection period and the second collection period may be preset, the whole body image of the client is collected in the first collection period, and the audio data of the client is collected in the second collection period.
  • the extraction module 202 is used to extract expression features and body features of predetermined points from the whole body image, and to extract speech rate features and pitch features from the audio data.
  • Facial expression is a form of body language that can complete fine information communication.
  • the human face has ten muscles, which can produce extremely rich expressions and accurately convey a variety of different mentalities and emotions.
  • the customer's affirmation and emotion can be determined through facial expressions. Emotions such as negation, pleasure and disappointment, satisfaction and dissatisfaction.
  • the key parts of satisfactory performance are the mouth, cheeks, eyebrows, and forehead, and the key parts of dissatisfied performance are mouth and brow.
  • the corners of the mouth, eyes, eyebrows, forehead, and chin can be preset as feature points, that is, the corners of the mouth, eyes, eyebrows, forehead, and chin are predetermined points.
  • the extracting the facial expression features of the predetermined points from the full-body image includes: detecting a human face from the full-body image according to a preset face detection algorithm; and extracting the facial expression features of the predetermined points in the human face .
  • Gestures, standing postures and body postures can show certain emotions, for example, clapping expresses excitement; pausing to express anger, rubbing hands expressing anxiety, and hammerhead express Depressed and so on. Therefore, after acquiring the full-body image of the customer, the body features such as gestures, standing posture, and body posture in the full-body image are extracted.
  • the speed of speech and the height of the pitch can express different emotions. For example, when a person is angry, the speed of speech is higher and the pitch is higher; when a person is happy, the speed of speech is slower and the pitch is moderate; In the case of sadness, the speech rate is slower and the pitch is lower. Therefore, after acquiring the audio data of the client's preset duration, the speech rate and pitch in the audio data are extracted.
  • the input module 203 is configured to input the expression characteristics into a pre-trained facial expression satisfaction recognition model, input the limb characteristics into the pre-trained limb satisfaction recognition model, and combine the speech rate characteristics and pitch The features are input into the pre-trained speech satisfaction recognition model.
  • the facial expression satisfaction recognition model, body satisfaction recognition model, and voice satisfaction recognition model are all pre-trained satisfaction recognition models. After the facial expression features, body features, speech rate features, and tone features are obtained At the same time, the expression characteristics, body characteristics, speech rate characteristics and tone characteristics are respectively input into the expression satisfaction recognition model, the body satisfaction recognition model, and the speech satisfaction recognition model for satisfaction recognition.
  • the training module 204 is used to train the facial expression satisfaction recognition model, including:
  • test pass rate is greater than or equal to the preset pass rate threshold, end the training of the facial expression satisfaction recognition model; otherwise, when the test pass rate is less than the preset pass rate threshold, retrain the expression Satisfaction recognition model until the test pass rate is greater than or equal to the preset pass rate threshold.
  • the training process of the body satisfaction recognition model and the speech satisfaction recognition model is the same as the facial expression satisfaction recognition process, and will not be elaborated here.
  • different satisfaction scores are preset according to different expression characteristics, body characteristics, speech speed characteristics, and tone characteristics. For example, a happy expression corresponds to a satisfaction score of 5 points, and an angry expression corresponds to satisfaction. The degree is -5 points. For ease of presentation, the satisfaction score corresponding to the expressionlessness is recorded as 0 points. The satisfaction score corresponding to fast speech and high pitch is -5 points; the satisfaction score corresponding to slow speech and moderate pitch is 5 points.
  • the user's facial expression characteristics, body characteristics, speech rate characteristics, tone characteristics, and satisfaction scores can be used as new data to increase the number of data sets, and retrain based on the new data sets Expression satisfaction recognition model, body satisfaction recognition model, and speech satisfaction recognition model, so as to continuously improve the recognition rate of each satisfaction recognition model.
  • the second acquisition module 205 is configured to acquire the first satisfaction score output by the facial expression satisfaction recognition model, the second satisfaction score output by the limb satisfaction recognition model, and the third satisfaction score output by the voice satisfaction recognition model. Satisfaction score.
  • the first satisfaction score can be output through the expression satisfaction recognition model, and the first satisfaction score represents the value corresponding to the customer’s expression characteristics. Satisfaction situation.
  • the second satisfaction score can be output through the body satisfaction recognition model, and the second satisfaction score represents the satisfaction degree corresponding to the client's body characteristics happening.
  • the third satisfaction score can be output through the speech satisfaction recognition model, and the third satisfaction score represents the customer’s speech rate feature and tone The satisfaction level corresponding to the feature. Different satisfaction scores represent different satisfaction situations.
  • the calculation module 206 is configured to calculate and output the final satisfaction degree of the customer according to the first satisfaction score, the second satisfaction score, and the third satisfaction score.
  • the first satisfaction score, the second satisfaction score, and the third satisfaction score are added and averaged to obtain the final satisfaction of the customer.
  • the customer's final satisfaction is calculated based on the expression characteristics, body characteristics, speech rate characteristics, and tone characteristics. It effectively combines the customer's all-round information, and the satisfaction obtained has more reference significance. And when the customer has no expression, no voice interaction, or no body, the satisfaction can still be calculated.
  • the calculation module 206 calculates the final satisfaction degree of the customer according to the first satisfaction score, the second satisfaction score, and the third satisfaction score includes:
  • the first final satisfaction degree, the second final satisfaction degree, and the third final satisfaction degree are added and averaged to obtain the final satisfaction degree.
  • the facial features can most intuitively express the customer’s emotions, and the physical features have a certain inertia, it can be preset that the first weight value corresponding to the facial features is the largest and the second weight value corresponding to the physical features is the smallest.
  • the third weight value corresponding to the speech rate feature and the tone feature is centered.
  • the sum of the first weight value, the second weight value, and the third weight value is 1.
  • the facial expression satisfaction recognition model can output 12 first satisfaction scores and 12 first satisfaction scores.
  • the 12 first satisfaction scores are added up and divided by 12 to get the first average satisfaction score.
  • the second average satisfaction score and the third average satisfaction score can be calculated.
  • the final satisfaction degree is calculated according to the first average satisfaction score and the first weight value, the second average satisfaction score and the second weight value, the third average satisfaction score and the third weight value.
  • the final satisfaction calculated using statistical methods represents an overall satisfaction in the service process.
  • the satisfaction identification device 20 also includes:
  • the alarm module 207 is used to determine whether the final satisfaction degree is less than the preset satisfaction threshold; if it is determined that the final satisfaction degree is less than the preset satisfaction threshold, an alarm is issued to the customer service according to the preset alarm mode.
  • the degree of satisfaction of the customer service process is divided into four levels, the first level: the customer is very satisfied; the second level: the customer is relatively satisfied; the third level: the customer is basically satisfied; the fourth level: the customer is very satisfied .
  • Different levels of satisfaction correspond to different satisfaction scores.
  • the satisfaction score corresponding to the first level is 15-20 points, the satisfaction score corresponding to the first level is 10-15 points; the third level corresponds to satisfaction
  • the degree score is 5-10 points, and the satisfaction score corresponding to the fourth level is 0-5 points.
  • Set a satisfaction threshold in advance for example, 5 points. When the final satisfaction is less than 5 points, it is determined that the customer is very dissatisfied, and the customer service is alerted according to the preset alerting method.
  • the preset alarm mode may be to display the alarm content on the display screen of the customer service; or send the alarm information by email or short message.
  • the final satisfaction is obtained, and when the final satisfaction is lower than the preset satisfaction threshold, the customer service is alerted, which helps to improve the customer's subsequent service quality.
  • the device 20 for identifying customer satisfaction based on micro-expression further includes:
  • the number of records increases by 1;
  • the full-body image and voice fragments of the customer are obtained from the first collection period, and the customer’s satisfaction scores in the first collection period are identified according to multiple satisfaction recognition models.
  • the number of records When there is a satisfaction score When the value is less than the preset satisfaction score, the number of records will be increased by 1.
  • the subsequent collection period if the number of records increases, when the number of records is greater than the threshold of the number of records, it indicates that in the process of customer service, the customer has clearly shown unsatisfactory emotions and unsatisfactory emotions. The number of times is too high. At this time, the customer service needs to be alerted, so that the customer service can improve the quality of service, and prevent the customer's dissatisfaction from erupting out of control.
  • the customer satisfaction identification device 20 also includes:
  • the screening module 208 is used to screen out the lowest first target satisfaction score from the first satisfaction score, filter the lowest second target satisfaction score from the second satisfaction score, The lowest third target satisfaction score is selected from the third satisfaction score; the first target satisfaction score, the second target satisfaction score, and the third target satisfaction score are identified Value of the target time node; extract a video clip of the preset duration corresponding to the target time node from the acquired video stream of the customer service; send the video clip to the customer service.
  • the video stream of the entire service of the customer service is also captured by the camera device.
  • the satisfaction score output by the satisfaction recognition model will also be very low.
  • the video clip corresponding to this moment is extracted and sent to the customer for viewing and analysis, which is convenient for improving behaviors such as improper speech during follow-up services. Improve service quality.
  • the micro-expression-based customer satisfaction recognition device described in this application can be applied in fields such as smart government affairs, so as to promote the development of smart cities.
  • This application collects customer’s facial expression characteristics, body characteristics, speech rate characteristics, and tone characteristics during the service process, and then uses multiple satisfaction recognition models to perform analysis on the expression characteristics, body characteristics, speech rate characteristics, and tone characteristics. Identify, get different satisfaction scores, and finally calculate the final satisfaction based on the different satisfaction scores.
  • the customer’s satisfaction cannot be calculated when the customer is expressionless during the entire service process, and by adopting multiple features, comprehensive consideration of the customer’s all-round information is adopted.
  • the calculated satisfaction is more realistic, which improves the success rate of satisfaction collection and ensures the accuracy of satisfaction collection.
  • the terminal 3 includes a memory 31, at least one processor 32, at least one communication bus 33, and a transceiver 34.
  • the structure of the terminal shown in FIG. 3 does not constitute a limitation of the embodiment of the present application. It may be a bus-type structure or a star structure. The terminal 3 may also include more More or less other hardware or software, or different component arrangements.
  • the terminal 3 includes a smart device that can automatically perform numerical calculation and/or information processing in accordance with pre-set or stored instructions.
  • Its hardware includes, but is not limited to, a microprocessor, an application specific integrated circuit, Programmable gate arrays, digital processors and embedded devices, etc.
  • the terminal 3 may also include client equipment.
  • the client equipment includes, but is not limited to, any electronic product that can interact with the client through a keyboard, a mouse, a remote control, a touch panel, or a voice control device, for example, a personal computer. Computers, tablets, smart phones, digital cameras, etc.
  • terminal 3 is only an example. If other existing or future electronic products can be adapted to this application, they should also be included in the protection scope of this application and included here by reference.
  • the memory 31 is used to store computer-readable instructions and various data, such as the micro-expression-based customer satisfaction recognition device 20 installed in the terminal 3, and is used during the operation of the terminal 3. Achieve high-speed, automatic completion of program or data access.
  • the memory 31 includes volatile and non-volatile memory, such as random access memory (Random Access Memory, RAM), read-only memory (Read-Only Memory, ROM), and programmable read-only memory (Programmable Read-Only).
  • PROM Erasable Programmable Read-Only Memory
  • EPROM Erasable Programmable Read-Only Memory
  • OTPROM Electronic Erasable Programmable Read-Only Memory
  • EEPROM Electrically-Erasable Programmable Read-Only Memory
  • CD-ROM Compact Disc Read-Only Memory
  • the computer-readable storage medium may be non-volatile or volatile.
  • the at least one processor 32 may be composed of integrated circuits, for example, may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits with the same function or different functions, including one Or a combination of multiple central processing units (CPU), microprocessors, digital processing chips, graphics processors, and various control chips.
  • the at least one processor 32 is the control core (Control Unit) of the terminal 3.
  • Various interfaces and lines are used to connect the various components of the entire terminal 3, and by running or executing programs or modules stored in the memory 31, And call the data stored in the memory 31 to perform various functions of the terminal 3 and process data, for example, perform the function of identifying customer satisfaction based on micro-expressions.
  • the at least one communication bus 33 is configured to implement connection and communication between the memory 31 and the at least one processor 32 and the like.
  • the terminal 3 may also include a power source (such as a battery) for supplying power to various components.
  • the power source may be logically connected to the at least one processor 32 through a power management device, so as to realize management through the power management device. Functions such as charging, discharging, and power management.
  • the power supply may also include any components such as one or more DC or AC power supplies, recharging devices, power failure detection circuits, power converters or inverters, and power status indicators.
  • the terminal 3 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.
  • the above-mentioned integrated unit implemented in the form of a software function module may be stored in a computer readable storage medium.
  • the above-mentioned software function module is stored in a storage medium and includes several instructions to make a computer device (which may be a personal computer, a terminal, or a network device, etc.) or a processor execute the method described in each embodiment of the present application. section.
  • the at least one processor 32 can execute the operating device of the terminal 3 and various installed applications (such as the micro-expression-based customer satisfaction recognition device 20) , Computer-readable instructions, etc., for example, the various modules mentioned above.
  • the memory 31 stores computer-readable instructions, and the at least one processor 32 can call the computer-readable instructions stored in the memory 31 to perform related functions.
  • the various modules described in FIG. 2 are computer-readable instructions stored in the memory 31 and executed by the at least one processor 32, so as to realize the functions of the various modules to achieve micro-expression-based The purpose of customer satisfaction identification.
  • the memory 31 stores a plurality of instructions, and the plurality of instructions are executed by the at least one processor 32 to realize the function of identifying customer satisfaction based on micro-expressions.
  • the disclosed device and method can be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the modules is only a logical function division, and there may be other division methods in actual implementation.
  • modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional modules in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit may be implemented in the form of hardware, or may be implemented in the form of hardware plus software functional modules.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Human Computer Interaction (AREA)
  • Theoretical Computer Science (AREA)
  • Psychiatry (AREA)
  • Business, Economics & Management (AREA)
  • Computational Linguistics (AREA)
  • Accounting & Taxation (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Hospice & Palliative Care (AREA)
  • Finance (AREA)
  • General Business, Economics & Management (AREA)
  • Signal Processing (AREA)
  • Economics (AREA)
  • Child & Adolescent Psychology (AREA)
  • Acoustics & Sound (AREA)
  • Marketing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Social Psychology (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Telephonic Communication Services (AREA)
  • Image Analysis (AREA)

Abstract

本申请涉及人工智能技术领域，提供了一种基于微表情的客户满意度识别方法、装置、终端及介质，所述方法包括：每隔预设采集周期获取客户的全身图像和预设时长的音频数据；从全身图像中提取预定点的表情特征和肢体特征，从音频数据中提取语速特征和音调特征；同时将表情特征输入至表情满意度识别模型中得到第一满意度分值，将肢体特征输入至肢体满意度识别模型中得到第二满意度分值，将语速特征和音调特征输入至语音满意度识别模型中得到第三满意度分值；根据所述第一、第二和第三满意度分值计算所述客户的最终满意度并输出。本申请能够解决在整个服务过程中客户无表情时无法计算客户的满意度的技术问题。

Description

基于微表情的客户满意度识别方法、装置、终端及介质
本申请要求于2019年08月26日提交中国专利局、申请号为201910792765.6，发明名称为“基于微表情的客户满意度识别方法、装置、终端及介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能技术领域，具体涉及一种基于微表情的客户满意度识别方法、装置、终端及介质。
背景技术
随着人类对服务质量的要求越来越高，客户对服务的满意度成为了人们关注的焦点，目前的满意度调查都是通过用户主动输入或填写问卷的形式获取，但通常情况下用户不愿意输入或者碍于情面不好意思输入较低的分数，导致满意度调查结果不准确。发明人意识到，现有技术中虽然也有通过微表情识别技术来识别用户的满意度，但是微表情识别技术需要获取到用户的面部表情，才能识别出用户的满意度，倘若用户无面部表情时，或者在用户佩戴有口罩、低头等情况下导致无法获取面部表情时，则无法通过微表情识别技术来识别用户的满意度。
因此，有必要提供一种新的方案，能够解决客户无表情时客户的满意度的识别问题。
发明内容
鉴于以上内容，有必要提出一种基于微表情的客户满意度识别方法、装置、终端及介质，能够解决在整个服务过程中客户无表情时无法计算客户的满意度的技术问题。
本申请的第一方面提供一种基于微表情的客户满意度识别方法，所述方法包括：
每隔预设采集周期获取客户的全身图像和预设时长的音频数据；
从所述全身图像中提取预定点的表情特征和肢体特征，及从所述音频数据中提取语速特征和音调特征；
同时将所述表情特征输入至预先训练好的表情满意度识别模型中，将所述肢体特征输入至预先训练好的肢体满意度识别模型中，将所述语速特征和音调特征输入至预先训练好的语音满意度识别模型中；
获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值；
根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度并输出。
本申请的第二方面提供一种基于微表情的客户满意度识别装置，所述装置包括：
第一获取模块，用于每隔预设采集周期获取客户的全身图像和预设时长的音频数据；
提取模块，用于从所述全身图像中提取预定点的表情特征和肢体特征，及从所述音频数据中提取语速特征和音调特征；
输入模块，用于同时将所述表情特征输入至预先训练好的表情满意度识别模型中，将所述肢体特征输入至预先训练好的肢体满意度识别模型中，将所述语速特征和音调特征输入至预先训练好的语音满意度识别模型中；
第二获取模块，用于获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值；
计算模块，用于根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度并输出。
本申请的第三方面提供一种终端，所述终端包括处理器，所述处理器用于执行存储器中 存储的计算机可读指令时实现以下步骤：
每隔预设采集周期获取客户的全身图像和预设时长的音频数据；
从所述全身图像中提取预定点的表情特征和肢体特征，及从所述音频数据中提取语速特征和音调特征；
同时将所述表情特征输入至预先训练好的表情满意度识别模型中，将所述肢体特征输入至预先训练好的肢体满意度识别模型中，将所述语速特征和音调特征输入至预先训练好的语音满意度识别模型中；
获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值；
根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度并输出。
本申请的第四方面提供一种计算机可读存储介质，所述计算机可读存储介质上存储有计算机可读指令，所述计算机可读指令被处理器执行时实现以下步骤：
每隔预设采集周期获取客户的全身图像和预设时长的音频数据；
从所述全身图像中提取预定点的表情特征和肢体特征，及从所述音频数据中提取语速特征和音调特征；
同时将所述表情特征输入至预先训练好的表情满意度识别模型中，将所述肢体特征输入至预先训练好的肢体满意度识别模型中，将所述语速特征和音调特征输入至预先训练好的语音满意度识别模型中；
获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值；
根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度并输出。
综上所述，本申请所述的基于微表情的客户满意度识别方法、装置、终端及介质，可应用在智慧政务等领域，从而推动智慧城市的发展。本申请通过在服务过程中，采集客户的表情特征、肢体特征、语速特征和音调特征，然后利用多个满意度识别模型分别对所述的表情特征、肢体特征、语速特征和音调特征进行识别，得到不同的满意度分值，最后基于不同的满意度分值计算得到最终的满意度。相对于现有技术中单一采用面部表情而言，能够解决在整个服务过程中客户无表情时无法计算客户的满意度的技术问题，且通过采用多个特征，综合考虑了客户的全方位信息，计算得到的满意度更具现实意义，提高了满意度的采集成功率，保证了满意度采集的准确性。
附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据提供的附图获得其他的附图。
图1是本申请实施例一提供的基于微表情的客户满意度识别方法的流程图。
图2是本申请实施例二提供的基于微表情的客户满意度识别装置的结构图。
图3是本申请实施例三提供的终端的结构示意图。
如下具体实施方式将结合上述附图进一步说明本申请。
具体实施方式
为了能够更清楚地理解本申请的上述目的、特征和优点，下面结合附图和具体实施例对本申请进行详细描述。需要说明的是，在不冲突的情况下，本申请的实施例及实施例中的特征可以相互组合。
在下面的描述中阐述了很多具体细节以便于充分理解本申请，所描述的实施例仅仅是本 申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
除非另有定义，本文所使用的所有的技术和科学术语与属于本申请的技术领域的技术人员通常理解的含义相同。本文中在本申请的说明书中所使用的术语只是为了描述具体的实施例的目的，不是旨在于限制本申请。
实施例一
图1是本申请实施例一提供的基于微表情的客户满意度识别方法的流程图。
在本实施例中，所述基于微表情的客户满意度识别方法可以应用于终端中，对于需要进行语音控制的终端，可以直接在终端上集成本申请的方法所提供的基于微表情的客户满意度识别的功能，或者以软件开发工具包(Software Development Kit，SKD)的形式运行在终端中。
如图1所示，所述基于微表情的客户满意度识别方法具体包括以下步骤，根据不同的需求，该流程图中步骤的顺序可以改变，某些可以省略。
S11，每隔预设采集周期获取客户的全身图像和预设时长的音频数据。
本实施例中，可以预先设置采集周期，如可以设置每5秒或者10秒为一个采集周期。
在客服与客户的交互过程中，控制预先装设的摄像装置每隔预设周期获取一次客户的全身图像，在获取客户的全身图像的同时，还采集客户的预设时长的一段音频数据。
在其他实施例中，还可以预先设置第一采集周期和第二采集周期，在所述第一采集周期内采集客户的全身图像，在所述第二采集周期内采集客户的音频数据。
S12，从所述全身图像中提取预定点的表情特征和肢体特征，及从所述音频数据中提取语速特征和音调特征。
面部表情是一种可完成精细信息沟通的体语形式，人的面部有十块肌肉，可产生极其丰富的表情，准确传达各种不同的心态和情感，通过面部表情可确定出客户的肯定和否定、愉悦和失望、满意和不满意等情感。一般来说，表现满意的关键部位是嘴、颊、眉、额头，表现不满意的关键部位是嘴、眉头。譬如，一个人眉毛上扬、挤在一起，呈现出的是一种恐惧、忧虑的表情；鼻孔外翻、嘴唇紧抿，呈现出的是一种愤怒的表情；下巴扬起、嘴角下垂，呈现出的是一种自责的表情。因此，可以预先设定嘴角、眼脸、眉毛、额头、下巴为特征点，即，嘴角、眼脸、眉毛、额头、下巴等部位为预定点。
具体的，所述从所述全身图像中提取预定点的表情特征包括：根据预先设置的人脸检测算法从所述全身图像中检测出人脸；提取出所述人脸中预定点的表情特征。
在通过肢体动作表达情绪时，也会有一些惯用的动作，手势、站姿和身体姿势都能表现出某种情绪，例如，鼓掌表示兴奋；顿足表示生气，搓手表示焦虑，锤头表示沮丧等。因此，在获取到客户的全身图像之后，提取出全身图像中的手势、站姿和身体姿势等肢体特征。
语速的快慢、音调的高低能够表达不同的情绪，比如，人在愤怒的情况下，语速较快、音调较高；人在欢喜的情况下，则语速较缓、音调适中；而人在悲伤的情况下，语速较慢、音调较低。因此，在获取到客户的预设时长的音频数据之后，提取出所述音频数据中的语速和音调。
本实施例中，提取表情特征、肢体特征、语速特征和音调特征均为现有技术，在此不再详细赘述。
S13，将所述表情特征输入至预先训练好的表情满意度识别模型中，将所述肢体特征输入至预先训练好的肢体满意度识别模型中，将所述语速特征和音调特征输入至预先训练好的语音满意度识别模型中。
本实施例中，所述表情满意度识别模型、肢体满意度识别模型、语音满意度识别模型均是预先训练好的满意度识别模型，在得到表情特征、肢体特征、语速特征和音调特征之后，同时将表情特征、肢体特征、语速特征和音调特征分别输入至表情满意度识别模型、肢体满意度识别模型、语音满意度识别模型中进行满意度识别。
其中，所述表情满意度识别模型的训练过程包括：
1)获取历史用户的表情特征及对应的满意度分值，形成数据集；
2)将所述数据集随机分为第一数量的训练集和第二数量的测试集；
3)将所述训练集输入至预设卷积神经网络中进行训练，得到表情满意度识别模型；
4)将所述测试集输入至所述表情满意度识别模型中进行测试，得到测试通过率；
5)判断所述测试通过率是否大于预设通过率阈值；
6)当所述测试通过率大于或者等于所述预设通过率阈值时，结束表情满意度识别模型的训练；否则，当所述测试通过率小于所述预设通过率阈值时，重新训练表情满意度识别模型直至所述测试通过率大于或者等于所述预设通过率阈值。
关于所述肢体满意度识别模型和语音满意度识别模型的训练过程同所述表情满意度识别过程，在此不再详细阐述。
本实施例中，根据不同的表情特征、肢体特征、语速特征和音调特征预先设置不同的满意度分值，例如，开心的表情对应的满意度分值为5分，恼羞成怒的表情对应的满意度为-5分。为了便于表述，将无表情对应的满意度分值记为0分。语速快、音调高对应的满意度分值为-5分；语速平缓、音调适中对应的满意度分值为5分。
可以在后续服务过程中，将用户的表情特征、肢体特征、语速特征和音调特征及满意度分值作为新的数据，以增加所述数据集的数量，并基于新的数据集来重新训练表情满意度识别模型、肢体满意度识别模型、语音满意度识别模型，从而不断的提高各个满意度识别模型的识别率。
S14，获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值。
本实施例中，将表情特征输入至表情满意度识别模型中之后，即可通过表情满意度识别模型输出第一满意度分值，所述第一满意度分值代表了客户的表情特征对应的满意度情况。同理，将肢体特征输入至肢体满意度识别模型中之后，即可通过肢体满意度识别模型输出第二满意度分值，所述第二满意度分值代表了客户的肢体特征对应的满意度情况。将语速特征和音调特征输入至语音满意度识别模型中之后，即可通过语音满意度识别模型输出第三满意度分值，所述第三满意度分值代表了客户的语速特征和音调特征对应的满意度情况。不同的满意度分值，代表了不同的满意度情况。
S15，根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度并输出。
本实施例中，将所述第一满意度分值、第二满意度分值和第三满意度分值进行加和平均即可得到客户的最终满意度。
因此可见，客户的最终满意度是根据表情特征、肢体特征、语速特征和音调特征综合计算得到的，有效的结合了客户的全方位信息，得到的满意度更具参考意义。且当客户无表情、或者无语音交互、或者无肢体中的任意一种的情况下，依然能计算得到满意度。
优选的，所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度包括：
统计采集的次数；
根据所述采集的次数计算多个所述第一满意度分值的第一平均满意度分值、多个所述第二满意度分值的第二平均满意度分值及多个所述第三满意度分值的第三平均满意度分值；
计算所述第一平均满意度分值与预设第一权重值的乘积，得到第一最终满意度；
计算所述第二平均满意度分值与预设第二权重值的乘积，得到第二最终满意度；
计算所述第三平均满意度分值与预设第三权重值的乘积，得到第三最终满意度；
对所述第一最终满意度、所述第二最终满意度和所述第三最终满意度进行加和平均，得到所述最终满意度。
本实施例中，由于表情特征最能直观的表达出客户的情绪，肢体特征带有一定的惯性， 因此，可以预先设置表情特征对应的第一权重值最大，肢体特征对应的第二权重值最小，语速特征和音调特征对应的第三权重值居中。所述第一权重值、第二权重值和第三权重值之和为1。
示例性，假设在2分钟的交互过程中，每隔10秒采集一次客户的全身图像和3秒的语音片段，则通过表情满意度识别模型可以输出12个第一满意度分值、12个第二满意度分值和12个第三满意度分值。将12个第一满意度分值进行加总后除以12，即可得到第一平均满意度分值。同理，可以计算出第二平均满意度分值及第三平均满意度分值。最后根据第一平均满意度分值与第一权重值、第二平均满意度分值与第二权重值、第三平均满意度分值与第三权重值，计算得到最终满意度。采用统计学的方法计算得到的最终满意度代表了在服务的过程中的一个总体的满意度。
进一步的，在所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度之后，所述方法还包括：
判断所述最终满意度是否小于预设满意度阈值；
若确定所述最终满意度小于所述预设满意度阈值时，根据预设告警方式对客服进行告警。
本实施例中，将客服的服务过程的满意程度分为四个等级，第一等级：客户非常满意；第二等级：客户比较满意；第三等级：客户基本满意；第四等级：客户非常满意。不同等级的满意度对应不同的满意度分值，如第一等级对应的满意度分值为15-20分，第一等级对应的满意度分值为10-15分；第三等级对应的满意度分值为5-10分，第四等级对应的满意度分值为0-5分。预先设置一个满意度阈值，例如，5分，当最终满意度小于5分时，确定客户非常不满意，根据预先设置的告警方式对客服进行告警。
所述预设告警方式可以是，在客服的显示屏幕上显示告警内容；或者通过邮件、短信方式发送告警信息。
通过在服务结束后，得到最终满意度，并当最终满意度低于预设满意度阈值时，对客服进行告警，有助于提高客服在后续的服务质量。
在其他实施例中，当获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值之后，所述方法还包括：
判断所述第一满意度分值是否小于第一满意度分值阈值、所述第二满意度分值是否小于第二满意度分值阈值、所述第三满意度分值是否小于第三满意度分值阈值；
当确定所述第一满意度分值小于所述第一满意度分值阈值，或者所述第二满意度分值小于所述第二满意度分值阈值，或者所述第三满意度分值小于所述第三满意度分值阈值时，记录次数增加1；
判断所述记录次数是否大于记录次数阈值；
当确定所述记录次数大于所述记录次数阈值时，根据所述预设告警方式对客户进行告警。
本实施例中，从第一个采集周期获取到客户的全身图像和语音片段，并根据多个满意度识别模型识别出客户这第一个采集周期的满意度分值，当有一个满意度分值小于预设满意度分值时，将记录次数增加1。在后续的采集周期内，若随着记录次数的增加，当记录次数大于记录次数阈值时，表明在客服服务的过程中，客户已经明显的表现出了不满意的情绪，且不满意的情绪的次数较多，此时需要对客服进行告警，使得客服提高服务质量，避免将客户的不满意情绪爆发到不可收拾的地步。
更进一步的，在所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度之后，所述方法还包括：
从所述第一满意度分值中筛选出最低的第一目标满意度分值、从所述第二满意度分值中筛选出最低的第二目标满意度分值、所述第三满意度分值中筛选出最低的第三目标满意度分值；
识别所述第一目标满意度分值、所述第二目标满意度分值、所述第三目标满意度分值的 目标时间节点；
从获取的客服的视频流中提取对应所述目标时间节点的预设时长的视频片段；
将所述视频片段发送给所述客服。
本实施例中，还通过摄像装置拍摄客服的整个服务的视频流，由于在客服与客户交互的过程中，可能会存在某一时刻言语不当或者其他因素导致在该时刻时，客户的满意度非常低，此时通过满意度识别模型输出的满意度分值也会非常低，将该时刻对应的视频片段提取出来发送给客户进行观看与分析，便于后续服务时改善言语不当等行为，有助于提高服务质量。
本申请所述的一种基于微表情的客户满意度识别方法，可应用在智慧政务等领域，从而推动智慧城市的发展。本申请通过在服务过程中，采集客户的表情特征、肢体特征、语速特征和音调特征，然后利用多个满意度识别模型分别对所述的表情特征、肢体特征、语速特征和音调特征进行识别，得到不同的满意度分值，最后基于不同的满意度分值计算得到最终的满意度。相对于现有技术中单一采用面部表情而言，能够解决在整个服务过程中客户无表情时无法计算客户的满意度的技术问题，且通过采用多个特征，综合考虑了客户的全方位信息，计算得到的满意度更具现实意义，提高了满意度的采集成功率，保证了满意度采集的准确性。
实施例二
图2是本申请实施例二提供的基于微表情的客户满意度识别装置的结构图。
在一些实施例中，所述基于微表情的客户满意度识别装置20可以包括多个由计算机可读指令所组成的功能模块。所述基于微表情的客户满意度识别装置20中的各个程序段的计算机可读指令可以存储于终端的存储器中，并由所述至少一个处理器所执行，以执行(详见图1描述)基于微表情的客户满意度识别的功能。
本实施例中，所述基于微表情的客户满意度识别装置20，根据其所执行的功能，可以被划分为多个功能模块。所述功能模块可以包括：第一获取模块201、提取模块202、输入模块203、训练模块204、第二获取模块205、计算模块206、告警模块207及筛选模块208。本申请所称的模块是指一种能够被至少一个处理器所执行并且能够完成固定功能的一系列计算机可读指令，其存储在存储器中。在本实施例中，关于各模块的功能将在后续的实施例中详述。
第一获取模块201，用于每隔预设采集周期获取客户的全身图像和预设时长的音频数据。
本实施例中，可以预先设置采集周期，如可以设置每5秒或者10秒为一个采集周期。
在客服与客户的交互过程中，控制预先装设的摄像装置每隔预设周期获取一次客户的全身图像，在获取客户的全身图像的同时，还采集客户的预设时长的一段音频数据。
在其他实施例中，还可以预先设置第一采集周期和第二采集周期，在所述第一采集周期内采集客户的全身图像，在所述第二采集周期内采集客户的音频数据。
提取模块202，用于从所述全身图像中提取预定点的表情特征和肢体特征，及从所述音频数据中提取语速特征和音调特征。
面部表情是一种可完成精细信息沟通的体语形式，人的面部有十块肌肉，可产生极其丰富的表情，准确传达各种不同的心态和情感，通过面部表情可确定出客户的肯定和否定、愉悦和失望、满意和不满意等情感。一般来说，表现满意的关键部位是嘴、颊、眉、额头，表现不满意的关键部位是嘴、眉头。譬如，一个人眉毛上扬、挤在一起，呈现出的是一种恐惧、忧虑的表情；鼻孔外翻、嘴唇紧抿，呈现出的是一种愤怒的表情；下巴扬起、嘴角下垂，呈现出的是一种自责的表情。因此，可以预先设定嘴角、眼脸、眉毛、额头、下巴为特征点，即，嘴角、眼脸、眉毛、额头、下巴等部位为预定点。
具体的，所述从所述全身图像中提取预定点的表情特征包括：根据预先设置的人脸检测算法从所述全身图像中检测出人脸；提取出所述人脸中预定点的表情特征。
在通过肢体动作表达情绪时，也会有一些惯用的动作，手势、站姿和身体姿势都能表现出某种情绪，例如，鼓掌表示兴奋；顿足表示生气，搓手表示焦虑，锤头表示沮丧等。因此，在获取到客户的全身图像之后，提取出全身图像中的手势、站姿和身体姿势等肢体特征。
语速的快慢、音调的高低能够表达不同的情绪，比如，人在愤怒的情况下，语速较快、音调较高；人在欢喜的情况下，则语速较缓、音调适中；而人在悲伤的情况下，语速较慢、音调较低。因此，在获取到客户的预设时长的音频数据之后，提取出所述音频数据中的语速和音调。
本实施例中，提取表情特征、肢体特征、语速特征和音调特征均为现有技术，在此不再详细赘述。
输入模块203，用于将所述表情特征输入至预先训练好的表情满意度识别模型中，将所述肢体特征输入至预先训练好的肢体满意度识别模型中，将所述语速特征和音调特征输入至预先训练好的语音满意度识别模型中。
本实施例中，所述表情满意度识别模型、肢体满意度识别模型、语音满意度识别模型均是预先训练好的满意度识别模型，在得到表情特征、肢体特征、语速特征和音调特征之后，同时将表情特征、肢体特征、语速特征和音调特征分别输入至表情满意度识别模型、肢体满意度识别模型、语音满意度识别模型中进行满意度识别。
训练模块204，用于训练所述表情满意度识别模型，包括：
1)获取历史用户的表情特征及对应的满意度分值，形成数据集；
2)将所述数据集随机分为第一数量的训练集和第二数量的测试集；
3)将所述训练集输入至预设卷积神经网络中进行训练，得到表情满意度识别模型；
4)将所述测试集输入至所述表情满意度识别模型中进行测试，得到测试通过率；
5)判断所述测试通过率是否大于预设通过率阈值；
6)当所述测试通过率大于或者等于所述预设通过率阈值时，结束表情满意度识别模型的训练；否则，当所述测试通过率小于所述预设通过率阈值时，重新训练表情满意度识别模型直至所述测试通过率大于或者等于所述预设通过率阈值。
关于所述肢体满意度识别模型和语音满意度识别模型的训练过程同所述表情满意度识别过程，在此不再详细阐述。
本实施例中，根据不同的表情特征、肢体特征、语速特征和音调特征预先设置不同的满意度分值，例如，开心的表情对应的满意度分值为5分，恼羞成怒的表情对应的满意度为-5分。为了便于表述，将无表情对应的满意度分值记为0分。语速快、音调高对应的满意度分值为-5分；语速平缓、音调适中对应的满意度分值为5分。
可以在后续服务过程中，将用户的表情特征、肢体特征、语速特征和音调特征及满意度分值作为新的数据，以增加所述数据集的数量，并基于新的数据集来重新训练表情满意度识别模型、肢体满意度识别模型、语音满意度识别模型，从而不断的提高各个满意度识别模型的识别率。
第二获取模块205，用于获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值。
本实施例中，将表情特征输入至表情满意度识别模型中之后，即可通过表情满意度识别模型输出第一满意度分值，所述第一满意度分值代表了客户的表情特征对应的满意度情况。同理，将肢体特征输入至肢体满意度识别模型中之后，即可通过肢体满意度识别模型输出第二满意度分值，所述第二满意度分值代表了客户的肢体特征对应的满意度情况。将语速特征和音调特征输入至语音满意度识别模型中之后，即可通过语音满意度识别模型输出第三满意度分值，所述第三满意度分值代表了客户的语速特征和音调特征对应的满意度情况。不同的满意度分值，代表了不同的满意度情况。
计算模块206，用于根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度并输出。
本实施例中，将所述第一满意度分值、第二满意度分值和第三满意度分值进行加和平均即可得到客户的最终满意度。
因此可见，客户的最终满意度是根据表情特征、肢体特征、语速特征和音调特征综合计 算得到的，有效的结合了客户的全方位信息，得到的满意度更具参考意义。且当客户无表情、或者无语音交互、或者无肢体中的任意一种的情况下，依然能计算得到满意度。
优选的，所述计算模块206根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度包括：
统计采集的次数；
根据所述采集的次数计算多个所述第一满意度分值的第一平均满意度分值、多个所述第二满意度分值的第二平均满意度分值及多个所述第三满意度分值的第三平均满意度分值；
计算所述第一平均满意度分值与预设第一权重值的乘积，得到第一最终满意度；
计算所述第二平均满意度分值与预设第二权重值的乘积，得到第二最终满意度；
计算所述第三平均满意度分值与预设第三权重值的乘积，得到第三最终满意度；
对所述第一最终满意度、所述第二最终满意度和所述第三最终满意度进行加和平均，得到所述最终满意度。
本实施例中，由于表情特征最能直观的表达出客户的情绪，肢体特征带有一定的惯性，因此，可以预先设置表情特征对应的第一权重值最大，肢体特征对应的第二权重值最小，语速特征和音调特征对应的第三权重值居中。所述第一权重值、第二权重值和第三权重值之和为1。
示例性，假设在2分钟的交互过程中，每隔10秒采集一次客户的全身图像和3秒的语音片段，则通过表情满意度识别模型可以输出12个第一满意度分值、12个第二满意度分值和12个第三满意度分值。将12个第一满意度分值进行加总后除以12，即可得到第一平均满意度分值。同理，可以计算出第二平均满意度分值及第三平均满意度分值。最后根据第一平均满意度分值与第一权重值、第二平均满意度分值与第二权重值、第三平均满意度分值与第三权重值，计算得到最终满意度。采用统计学的方法计算得到的最终满意度代表了在服务的过程中的一个总体的满意度。
进一步的，在所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度之后，所述基于微表情的客户满意度识别装置20还包括：
告警模块207，用于判断所述最终满意度是否小于预设满意度阈值；若确定所述最终满意度小于所述预设满意度阈值时，根据预设告警方式对客服进行告警。
本实施例中，将客服的服务过程的满意程度分为四个等级，第一等级：客户非常满意；第二等级：客户比较满意；第三等级：客户基本满意；第四等级：客户非常满意。不同等级的满意度对应不同的满意度分值，如第一等级对应的满意度分值为15-20分，第一等级对应的满意度分值为10-15分；第三等级对应的满意度分值为5-10分，第四等级对应的满意度分值为0-5分。预先设置一个满意度阈值，例如，5分，当最终满意度小于5分时，确定客户非常不满意，根据预先设置的告警方式对客服进行告警。
所述预设告警方式可以是，在客服的显示屏幕上显示告警内容；或者通过邮件、短信方式发送告警信息。
通过在服务结束后，得到最终满意度，并当最终满意度低于预设满意度阈值时，对客服进行告警，有助于提高客服在后续的服务质量。
在其他实施例中，当获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值之后，所述基于微表情的客户满意度识别装置20还包括：
判断所述第一满意度分值是否小于第一满意度分值阈值、所述第二满意度分值是否小于第二满意度分值阈值、所述第三满意度分值是否小于第三满意度分值阈值；
当确定所述第一满意度分值小于所述第一满意度分值阈值，或者所述第二满意度分值小于所述第二满意度分值阈值，或者所述第三满意度分值小于所述第三满意度分值阈值时，记录次数增加1；
判断所述记录次数是否大于记录次数阈值；
当确定所述记录次数大于所述记录次数阈值时，根据所述预设告警方式对客户进行告警。
本实施例中，从第一个采集周期获取到客户的全身图像和语音片段，并根据多个满意度识别模型识别出客户这第一个采集周期的满意度分值，当有一个满意度分值小于预设满意度分值时，将记录次数增加1。在后续的采集周期内，若随着记录次数的增加，当记录次数大于记录次数阈值时，表明在客服服务的过程中，客户已经明显的表现出了不满意的情绪，且不满意的情绪的次数较多，此时需要对客服进行告警，使得客服提高服务质量，避免将客户的不满意情绪爆发到不可收拾的地步。
更进一步的，在所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度之后，所述基于微表情的客户满意度识别装置20还包括：
筛选模块208，用于从所述第一满意度分值中筛选出最低的第一目标满意度分值、从所述第二满意度分值中筛选出最低的第二目标满意度分值、所述第三满意度分值中筛选出最低的第三目标满意度分值；识别所述第一目标满意度分值、所述第二目标满意度分值、所述第三目标满意度分值的目标时间节点；从获取的客服的视频流中提取对应所述目标时间节点的预设时长的视频片段；将所述视频片段发送给所述客服。
本实施例中，还通过摄像装置拍摄客服的整个服务的视频流，由于在客服与客户交互的过程中，可能会存在某一时刻言语不当或者其他因素导致在该时刻时，客户的满意度非常低，此时通过满意度识别模型输出的满意度分值也会非常低，将该时刻对应的视频片段提取出来发送给客户进行观看与分析，便于后续服务时改善言语不当等行为，有助于提高服务质量。
本申请所述的一种基于微表情的客户满意度识别装置，可应用在智慧政务等领域，从而推动智慧城市的发展。本申请通过在服务过程中，采集客户的表情特征、肢体特征、语速特征和音调特征，然后利用多个满意度识别模型分别对所述的表情特征、肢体特征、语速特征和音调特征进行识别，得到不同的满意度分值，最后基于不同的满意度分值计算得到最终的满意度。相对于现有技术中单一采用面部表情而言，能够解决在整个服务过程中客户无表情时无法计算客户的满意度的技术问题，且通过采用多个特征，综合考虑了客户的全方位信息，计算得到的满意度更具现实意义，提高了满意度的采集成功率，保证了满意度采集的准确性。
实施例三
参阅图3所示，为本申请实施例三提供的终端的结构示意图。在本申请较佳实施例中，所述终端3包括存储器31、至少一个处理器32、至少一条通信总线33及收发器34。
本领域技术人员应该了解，图3示出的终端的结构并不构成本申请实施例的限定，既可以是总线型结构，也可以是星形结构，所述终端3还可以包括比图示更多或更少的其他硬件或者软件，或者不同的部件布置。
在一些实施例中，所述终端3包括一种能够按照事先设定或存储的指令，自动进行数值计算和/或信息处理的智能设备，其硬件包括但不限于微处理器、专用集成电路、可编程门阵列、数字处理器及嵌入式设备等。所述终端3还可包括客户设备，所述客户设备包括但不限于任何一种可与客户通过键盘、鼠标、遥控器、触摸板或声控设备等方式进行人机交互的电子产品，例如，个人计算机、平板电脑、智能手机、数码相机等。
需要说明的是，所述终端3仅为举例，其他现有的或今后可能出现的电子产品如可适应于本申请，也应包含在本申请的保护范围以内，并以引用方式包含于此。
在一些实施例中，所述存储器31用于存储计算机可读指令和各种数据，例如安装在所述终端3中的基于微表情的客户满意度识别装置20，并在终端3的运行过程中实现高速、自动地完成程序或数据的存取。所述存储器31包括易失性和非易失性存储器，例如随机存取存储器(Random Access Memory，RAM)、只读存储器(Read-Only Memory，ROM)、可编程只读存储器(Programmable Read-Only Memory，PROM)、可擦除可编程只读存储器(Erasable Programmable Read-Only Memory，EPROM)、一次可编程只读存储器(One-time Programmable Read-Only Memory，OTPROM)、电子擦除式可复写只读存储器(Electrically-Erasable  Programmable Read-Only Memory，EEPROM)、只读光盘(Compact Disc Read-Only Memory，CD-ROM)或其他光盘存储器、磁盘存储器、磁带存储器、或者其他能够用于携带或存储数据的计算机可读的存储介质。所述计算机可读存储介质可以是非易失性，也可以是易失性的。在一些实施例中，所述至少一个处理器32可以由集成电路组成，例如可以由单个封装的集成电路所组成，也可以是由多个相同功能或不同功能封装的集成电路所组成，包括一个或者多个中央处理器(Central Processing unit，CPU)、微处理器、数字处理芯片、图形处理器及各种控制芯片的组合等。所述至少一个处理器32是所述终端3的控制核心(Control Unit)，利用各种接口和线路连接整个终端3的各个部件，通过运行或执行存储在所述存储器31内的程序或者模块，以及调用存储在所述存储器31内的数据，以执行终端3的各种功能和处理数据，例如执行基于微表情的客户满意度识别的功能。
在一些实施例中，所述至少一条通信总线33被设置为实现所述存储器31以及所述至少一个处理器32等之间的连接通信。
尽管未示出，所述终端3还可以包括给各个部件供电的电源(比如电池)，优选的，电源可以通过电源管理装置与所述至少一个处理器32逻辑相连，从而通过电源管理装置实现管理充电、放电、以及功耗管理等功能。电源还可以包括一个或一个以上的直流或交流电源、再充电装置、电源故障检测电路、电源转换器或者逆变器、电源状态指示器等任意组件。所述终端3还可以包括多种传感器、蓝牙模块、Wi-Fi模块等，在此不再赘述。
应该了解，所述实施例仅为说明之用，在专利申请范围上并不受此结构的限制。
上述以软件功能模块的形式实现的集成的单元，可以存储在一个计算机可读取存储介质中。上述软件功能模块存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，终端，或者网络设备等)或处理器(processor)执行本申请各个实施例所述方法的部分。
在进一步的实施例中，结合图2，所述至少一个处理器32可执行所述终端3的操作装置以及安装的各类应用程序(如所述的基于微表情的客户满意度识别装置20)、计算机可读指令等，例如，上述的各个模块。
所述存储器31中存储有计算机可读指令，且所述至少一个处理器32可调用所述存储器31中存储的计算机可读指令以执行相关的功能。例如，图2中所述的各个模块是存储在所述存储器31中的计算机可读指令，并由所述至少一个处理器32所执行，从而实现所述各个模块的功能以达到基于微表情的客户满意度识别的目的。
在本申请的一个实施例中，所述存储器31存储多个指令，所述多个指令被所述至少一个处理器32所执行以实现基于微表情的客户满意度识别的功能。
具体地，所述至少一个处理器32对上述指令的具体实现方法可参考图1对应实施例中相关步骤的描述，在此不赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述模块的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式。
所述作为分离部件说明的模块可以是或者也可以不是物理上分开的，作为模块显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能模块可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用硬件加软件功能模块的形式实现。
对于本领域技术人员而言，显然本申请不限于上述示范性实施例的细节，而且在不背离本申请的精神或基本特征的情况下，能够以其他的具体形式实现本申请。因此，无论从哪一点来看，均应将实施例看作是示范性的，而且是非限制性的，本申请的范围由所附权利要求而不是上述说明限定，因此旨在将落在权利要求的等同要件的含义和范围内的所有变化涵括 在本申请内。不应将权利要求中的任何附图标记视为限制所涉及的权利要求。此外，显然“包括”一词不排除其他单元或，单数不排除复数。装置权利要求中陈述的多个单元或装置也可以由一个单元或装置通过软件或者硬件来实现。第一，第二等词语用来表示名称，而并不表示任何特定的顺序。
最后应说明的是，以上实施例仅用以说明本申请的技术方案而非限制，尽管参照较佳实施例对本申请进行了详细说明，本领域的普通技术人员应当理解，可以对本申请的技术方案进行修改或等同替换，而不脱离本申请技术方案的精神和范围。

Claims (20)

  1. 一种基于微表情的客户满意度识别方法，其中，所述方法包括：
    每隔预设采集周期获取客户的全身图像和预设时长的音频数据；
    从所述全身图像中提取预定点的表情特征和肢体特征，及从所述音频数据中提取语速特征和音调特征；
    同时将所述表情特征输入至预先训练好的表情满意度识别模型中，将所述肢体特征输入至预先训练好的肢体满意度识别模型中，将所述语速特征和音调特征输入至预先训练好的语音满意度识别模型中；
    获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值；
    根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度并输出。
  2. 如权利要求1所述的基于微表情的客户满意度识别方法，其中，所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度包括：
    统计采集的次数；
    根据所述采集的次数计算多个所述第一满意度分值的第一平均满意度分值、多个所述第二满意度分值的第二平均满意度分值及多个所述第三满意度分值的第三平均满意度分值；
    计算所述第一平均满意度分值与预设第一权重值的乘积，得到第一最终满意度；
    计算所述第二平均满意度分值与预设第二权重值的乘积，得到第二最终满意度；
    计算所述第三平均满意度分值与预设第三权重值的乘积，得到第三最终满意度；
    对所述第一最终满意度、所述第二最终满意度和所述第三最终满意度进行加和平均，得到所述最终满意度。
  3. 如权利要求1所述的基于微表情的客户满意度识别方法，其中，在所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度之后，所述方法还包括：
    判断所述最终满意度是否小于预设满意度阈值；
    若确定所述最终满意度小于所述预设满意度阈值时，根据预设告警方式对客服进行告警。
  4. 如权利要求3所述的基于微表情的客户满意度识别方法，其中，当获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值之后，所述方法还包括：
    判断所述第一满意度分值是否小于第一满意度分值阈值、所述第二满意度分值是否小于第二满意度分值阈值、所述第三满意度分值是否小于第三满意度分值阈值；
    当确定所述第一满意度分值小于所述第一满意度分值阈值，或者所述第二满意度分值小于所述第二满意度分值阈值，或者所述第三满意度分值小于所述第三满意度分值阈值时，记录次数增加1；
    判断所述记录次数是否大于记录次数阈值；
    当确定所述记录次数大于所述记录次数阈值时，根据所述预设告警方式对客户进行告警。
  5. 如权利要求1所述的基于微表情的客户满意度识别方法，其中，在所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度之后，所述方法还包括：
    从所述第一满意度分值中筛选出最低的第一目标满意度分值、从所述第二满意度分值中筛选出最低的第二目标满意度分值、所述第三满意度分值中筛选出最低的第三目标满意度分值；
    识别所述第一目标满意度分值、所述第二目标满意度分值、所述第三目标满意度分值的目标时间节点；
    从获取的客服的视频流中提取对应所述目标时间节点的预设时长的视频片段；
    将所述视频片段发送给所述客服。
  6. 如权利要求1所述的基于微表情的客户满意度识别方法，其中，所述表情满意度识别模型的训练过程包括：
    获取历史用户的表情特征及对应的满意度分值，形成数据集；
    将所述数据集随机分为第一数量的训练集和第二数量的测试集；
    将所述训练集输入至预设卷积神经网络中进行训练，得到表情满意度识别模型；
    将所述测试集输入至所述表情满意度识别模型中进行测试，得到测试通过率；
    判断所述测试通过率是否大于预设通过率阈值；
    当所述测试通过率大于或者等于所述预设通过率阈值时，结束表情满意度识别模型的训练；否则，当所述测试通过率小于所述预设通过率阈值时，重新训练表情满意度识别模型直至所述测试通过率大于或者等于所述预设通过率阈值。
  7. 如权利要求1至6中任意一项所述的基于微表情的客户满意度识别方法，其中，从所述全身图像中提取预定点的表情特征包括：
    根据预先设置的人脸检测算法从所述全身图像中检测出人脸；
    提取所述人脸中预定点的表情特征，所述预定点包括：嘴角、眼脸、眉毛、额头、下巴。
  8. 一种基于微表情的客户满意度识别装置，其中，所述装置包括：
    第一获取模块，用于每隔预设采集周期获取客户的全身图像和预设时长的音频数据；
    提取模块，用于从所述全身图像中提取预定点的表情特征和肢体特征，及从所述音频数据中提取语速特征和音调特征；
    输入模块，用于同时将所述表情特征输入至预先训练好的表情满意度识别模型中，将所述肢体特征输入至预先训练好的肢体满意度识别模型中，将所述语速特征和音调特征输入至预先训练好的语音满意度识别模型中；
    第二获取模块，用于获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值；
    计算模块，用于根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度并输出。
  9. 一种终端，其中，所述终端包括处理器，所述处理器用于执行存储器中存储的计算机可读指令以实现以下步骤：
    每隔预设采集周期获取客户的全身图像和预设时长的音频数据；
    从所述全身图像中提取预定点的表情特征和肢体特征，及从所述音频数据中提取语速特征和音调特征；
    同时将所述表情特征输入至预先训练好的表情满意度识别模型中，将所述肢体特征输入至预先训练好的肢体满意度识别模型中，将所述语速特征和音调特征输入至预先训练好的语音满意度识别模型中；
    获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值；
    根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度并输出。
  10. 如权利要求9所述的终端，其中，所述处理器执行所述计算机可读指令以实现根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度时，具体包括：
    统计采集的次数；
    根据所述采集的次数计算多个所述第一满意度分值的第一平均满意度分值、多个所述第二满意度分值的第二平均满意度分值及多个所述第三满意度分值的第三平均满意度分值；
    计算所述第一平均满意度分值与预设第一权重值的乘积，得到第一最终满意度；
    计算所述第二平均满意度分值与预设第二权重值的乘积，得到第二最终满意度；
    计算所述第三平均满意度分值与预设第三权重值的乘积，得到第三最终满意度；
    对所述第一最终满意度、所述第二最终满意度和所述第三最终满意度进行加和平均，得到所述最终满意度。
  11. 如权利要求9所述的终端，其中，在所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度之后，所述处理器执行所述计算机可读指令还用以实现以下步骤：
    判断所述最终满意度是否小于预设满意度阈值；
    若确定所述最终满意度小于所述预设满意度阈值时，根据预设告警方式对客服进行告警。
  12. 如权利要求11所述的终端，其中，当获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值之后，所述处理器执行所述计算机可读指令还用以实现以下步骤：
    判断所述第一满意度分值是否小于第一满意度分值阈值、所述第二满意度分值是否小于第二满意度分值阈值、所述第三满意度分值是否小于第三满意度分值阈值；
    当确定所述第一满意度分值小于所述第一满意度分值阈值，或者所述第二满意度分值小于所述第二满意度分值阈值，或者所述第三满意度分值小于所述第三满意度分值阈值时，记录次数增加1；
    判断所述记录次数是否大于记录次数阈值；
    当确定所述记录次数大于所述记录次数阈值时，根据所述预设告警方式对客户进行告警。
  13. 如权利要求9所述的终端，其中，在所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度之后，所述处理器执行所述计算机可读指令还用以实现以下步骤：
    从所述第一满意度分值中筛选出最低的第一目标满意度分值、从所述第二满意度分值中筛选出最低的第二目标满意度分值、所述第三满意度分值中筛选出最低的第三目标满意度分值；
    识别所述第一目标满意度分值、所述第二目标满意度分值、所述第三目标满意度分值的目标时间节点；
    从获取的客服的视频流中提取对应所述目标时间节点的预设时长的视频片段；
    将所述视频片段发送给所述客服。
  14. 如权利要求9所述的终端，其中，所述表情满意度识别模型的训练过程包括：
    获取历史用户的表情特征及对应的满意度分值，形成数据集；
    将所述数据集随机分为第一数量的训练集和第二数量的测试集；
    将所述训练集输入至预设卷积神经网络中进行训练，得到表情满意度识别模型；
    将所述测试集输入至所述表情满意度识别模型中进行测试，得到测试通过率；
    判断所述测试通过率是否大于预设通过率阈值；
    当所述测试通过率大于或者等于所述预设通过率阈值时，结束表情满意度识别模型的训练；否则，当所述测试通过率小于所述预设通过率阈值时，重新训练表情满意度识别模型直至所述测试通过率大于或者等于所述预设通过率阈值。
  15. 如权利要求9至14中任意一项所述的终端，其中，所述处理器执行所述计算机可读指令以实现从所述全身图像中提取预定点的表情特征时，具体包括：
    根据预先设置的人脸检测算法从所述全身图像中检测出人脸；
    提取所述人脸中预定点的表情特征，所述预定点包括：嘴角、眼脸、眉毛、额头、下巴。
  16. 一种计算机可读存储介质，所述计算机可读存储介质上存储有计算机可读指令，其中，所述计算机可读指令被处理器执行时实现以下步骤：
    每隔预设采集周期获取客户的全身图像和预设时长的音频数据；
    从所述全身图像中提取预定点的表情特征和肢体特征，及从所述音频数据中提取语速特征和音调特征；
    同时将所述表情特征输入至预先训练好的表情满意度识别模型中，将所述肢体特征输入至预先训练好的肢体满意度识别模型中，将所述语速特征和音调特征输入至预先训练好的语音满意度识别模型中；
    获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值；
    根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度并输出。
  17. 如权利要求16所述的计算机可读存储介质，其中，所述计算机可读指令被处理器执行以实现根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度时，具体包括：
    统计采集的次数；
    根据所述采集的次数计算多个所述第一满意度分值的第一平均满意度分值、多个所述第二满意度分值的第二平均满意度分值及多个所述第三满意度分值的第三平均满意度分值；
    计算所述第一平均满意度分值与预设第一权重值的乘积，得到第一最终满意度；
    计算所述第二平均满意度分值与预设第二权重值的乘积，得到第二最终满意度；
    计算所述第三平均满意度分值与预设第三权重值的乘积，得到第三最终满意度；
    对所述第一最终满意度、所述第二最终满意度和所述第三最终满意度进行加和平均，得到所述最终满意度。
  18. 如权利要求16所述的计算机可读存储介质，其中，在所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度之后，所述计算机可读指令被所述处理器执行还用以实现以下步骤：
    判断所述最终满意度是否小于预设满意度阈值；
    若确定所述最终满意度小于所述预设满意度阈值时，根据预设告警方式对客服进行告警。
  19. 如权利要求18所述的计算机可读存储介质，其中，当获取所述表情满意度识别模型输出的第一满意度分值、所述肢体满意度识别模型输出的第二满意度分值及语音满意度识别模型输出的第三满意度分值之后，所述计算机可读指令被所述处理器执行还用以实现以下步骤：
    判断所述第一满意度分值是否小于第一满意度分值阈值、所述第二满意度分值是否小于第二满意度分值阈值、所述第三满意度分值是否小于第三满意度分值阈值；
    当确定所述第一满意度分值小于所述第一满意度分值阈值，或者所述第二满意度分值小于所述第二满意度分值阈值，或者所述第三满意度分值小于所述第三满意度分值阈值时，记录次数增加1；
    判断所述记录次数是否大于记录次数阈值；
    当确定所述记录次数大于所述记录次数阈值时，根据所述预设告警方式对客户进行告警。
  20. 如权利要求16所述的计算机可读存储介质，其中，在所述根据所述第一满意度分值、所述第二满意度分值和所述第三满意度分值计算所述客户的最终满意度之后，所述计算机可读指令被所述处理器执行还用以实现以下步骤：
    从所述第一满意度分值中筛选出最低的第一目标满意度分值、从所述第二满意度分值中筛选出最低的第二目标满意度分值、所述第三满意度分值中筛选出最低的第三目标满意度分值；
    识别所述第一目标满意度分值、所述第二目标满意度分值、所述第三目标满意度分值的目标时间节点；
    从获取的客服的视频流中提取对应所述目标时间节点的预设时长的视频片段；
    将所述视频片段发送给所述客服。
PCT/CN2020/105631 2019-08-26 2020-07-29 基于微表情的客户满意度识别方法、装置、终端及介质 WO2021036664A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910792765.6A CN110705349A (zh) 2019-08-26 2019-08-26 基于微表情的客户满意度识别方法、装置、终端及介质
CN201910792765.6 2019-08-26

Publications (1)

Publication Number Publication Date
WO2021036664A1 true WO2021036664A1 (zh) 2021-03-04

Family

ID=69193479

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/105631 WO2021036664A1 (zh) 2019-08-26 2020-07-29 基于微表情的客户满意度识别方法、装置、终端及介质

Country Status (2)

Country Link
CN (1) CN110705349A (zh)
WO (1) WO2021036664A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113869773A (zh) * 2021-10-13 2021-12-31 北京卓思天成数据咨询股份有限公司 一种潜客满意度的测量方法及装置
CN117575548A (zh) * 2024-01-17 2024-02-20 华安证券股份有限公司 基于业务需求的服务方案智能化生成方法及系统

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110705349A (zh) * 2019-08-26 2020-01-17 深圳壹账通智能科技有限公司 基于微表情的客户满意度识别方法、装置、终端及介质
CN111460112A (zh) * 2020-03-02 2020-07-28 深圳壹账通智能科技有限公司 在线客服咨询方法、装置、介质及电子设备
CN111639518A (zh) * 2020-03-06 2020-09-08 上海卓繁信息技术股份有限公司 一种满意度测评方法，装置及存储设备
CN112329586A (zh) * 2020-10-30 2021-02-05 中国平安人寿保险股份有限公司 基于情绪识别的客户回访方法、装置及计算机设备
CN113128452A (zh) * 2021-04-30 2021-07-16 重庆锐云科技有限公司 一种基于图像识别的绿化满意度采集方法和系统
CN113177114B (zh) * 2021-05-28 2022-10-21 重庆电子工程职业学院 一种基于深度学习的自然语言语义理解方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109190550A (zh) * 2018-08-29 2019-01-11 沈阳康泰电子科技股份有限公司 联合微表情多输入信息的深度神经网络多源数据融合方法
CN109815803A (zh) * 2018-12-18 2019-05-28 平安科技（深圳）有限公司 面审风险控制方法、装置、计算机设备和存储介质
CN109858379A (zh) * 2019-01-03 2019-06-07 深圳壹账通智能科技有限公司 笑容真诚度检测方法、装置、存储介质和电子设备
CN110705349A (zh) * 2019-08-26 2020-01-17 深圳壹账通智能科技有限公司 基于微表情的客户满意度识别方法、装置、终端及介质

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10166091B2 (en) * 2014-02-21 2019-01-01 Trispera Dental Inc. Augmented reality dental design method and system
CN107705807B (zh) * 2017-08-24 2019-08-27 平安科技（深圳）有限公司 基于情绪识别的语音质检方法、装置、设备及存储介质
CN109858405A (zh) * 2019-01-17 2019-06-07 深圳壹账通智能科技有限公司 基于微表情的满意度评价方法、装置、设备及存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109190550A (zh) * 2018-08-29 2019-01-11 沈阳康泰电子科技股份有限公司 联合微表情多输入信息的深度神经网络多源数据融合方法
CN109815803A (zh) * 2018-12-18 2019-05-28 平安科技（深圳）有限公司 面审风险控制方法、装置、计算机设备和存储介质
CN109858379A (zh) * 2019-01-03 2019-06-07 深圳壹账通智能科技有限公司 笑容真诚度检测方法、装置、存储介质和电子设备
CN110705349A (zh) * 2019-08-26 2020-01-17 深圳壹账通智能科技有限公司 基于微表情的客户满意度识别方法、装置、终端及介质

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113869773A (zh) * 2021-10-13 2021-12-31 北京卓思天成数据咨询股份有限公司 一种潜客满意度的测量方法及装置
CN117575548A (zh) * 2024-01-17 2024-02-20 华安证券股份有限公司 基于业务需求的服务方案智能化生成方法及系统
CN117575548B (zh) * 2024-01-17 2024-03-22 华安证券股份有限公司 基于业务需求的服务方案智能化生成方法及系统

Also Published As

Publication number Publication date
CN110705349A (zh) 2020-01-17

Similar Documents

Publication Publication Date Title
WO2021036664A1 (zh) 基于微表情的客户满意度识别方法、装置、终端及介质
Kim et al. Towards emotionally aware AI smart classroom: Current issues and directions for engineering and education
US11010645B2 (en) Interactive artificial intelligence analytical system
Pampouchidou et al. Depression assessment by fusing high and low level features from audio, video, and text
CN111563487B (zh) 基于姿态识别模型的舞蹈评分方法及相关设备
US9875445B2 (en) Dynamic hybrid models for multimodal analysis
US20140212853A1 (en) Multi-modal modeling of temporal interaction sequences
US20140212854A1 (en) Multi-modal modeling of temporal interaction sequences
CN110874716A (zh) 面试测评方法、装置、电子设备及存储介质
US20180129647A1 (en) Systems and methods for dynamically collecting and evaluating potential imprecise characteristics for creating precise characteristics
Al Osman et al. Multimodal affect recognition: Current approaches and challenges
KR102276415B1 (ko) 개인 관심상황 발생 예측/인지 장치 및 방법
CN113591489B (zh) 语音交互方法、装置及相关设备
US20210201696A1 (en) Automated speech coaching systems and methods
Kim et al. Emotion spotting: Discovering regions of evidence in audio-visual emotion expressions
TW202117616A (zh) 適性化職缺媒合系統及方法
US11443554B2 (en) Determining and presenting user emotion
US10902301B2 (en) Information processing device and non-transitory computer readable medium storing information processing program
CN107463684A (zh) 语音回复方法及装置、计算机装置和计算机可读存储介质
Healy et al. Detecting demeanor for healthcare with machine learning
Zeng et al. Emotion recognition based on multimodal information
CN113468983A (zh) 一种情绪分析方法、装置、设备及存储介质
CN112686232A (zh) 基于微表情识别的教学评价方法、装置、电子设备及介质
CN112365956A (zh) 基于虚拟现实的心理治疗方法、装置、服务器及存储介质
Chiba et al. User modeling by using bag-of-behaviors for building a dialog system sensitive to the interlocutor’s internal state

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: 20858687

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 01.07.2022)

122 Ep: pct application non-entry in european phase

Ref document number: 20858687

Country of ref document: EP

Kind code of ref document: A1