Intelligent magic mirror personalized interaction method and system fusing voiceprint recognition

Abstract

The application discloses a kind of intelligent magic mirror personalized interaction method and system of fusion voiceprint identification, the method includes: obtaining target user current interaction data of target user;Target user historical interaction data corresponding to target user is obtained;Target user current interaction data and target user historical interaction data are compared with multidimensional similarity, and target user abnormal interaction index of target user current interaction is calculated;Whether target user abnormal interaction index is less than target user preset abnormal interaction index threshold is judged;If yes, according to target user current interaction data and target user historical normal personalized interaction data, target intelligent magic mirror personalized control instruction is generated;If not, it is judged that the voice instruction corresponding to target user current voice feature vector is target high-risk abnormal instruction, and anti-fraud early warning program is triggered.The application can effectively prevent AI voice imitation and recording attack and other fraudulent behaviors, protect user property and privacy security.

Description

Technical Field

[0001] This invention relates to the technical field of smart homes, and in particular to a personalized interaction method and system for a smart mirror that integrates voiceprint recognition. Background Technology

[0002] Current technologies for smart mirrors have several limitations that negatively impact user experience. First, voiceprint recognition suffers from poor environmental adaptability. While existing technologies achieve speech recognition and voiceprint extraction, they fail to consider the impact of environmental factors (temperature, humidity, noise) on voiceprint recognition. In practice, high humidity in bathrooms and kitchen fumes and noise can both reduce recognition rates. Second, they lack adaptability to individual states. Although current technologies can identify users and provide personalized information based on voiceprints, they don't account for the impact of changes in a user's condition, such as colds, fatigue, or emotional fluctuations, on voiceprint characteristics. Third, there's a lack of anomaly detection mechanisms. Current technologies assume all collected voiceprint data is normal, making it particularly vulnerable to AI voice imitation fraud, recording attacks, imitation by others, and user health abnormalities. Fourth, personalization is limited. Personalization is based solely on user identity, failing to incorporate real-time environment, geographical location, and user status. Finally, current smart mirrors offer limited interaction methods (mostly touch or simple voice control). Voiceprint recognition is only used for identity verification, lacking environmental integration and deep personalization capabilities. In addition, current AI voice synthesis technology (such as Deepfake voice) can be so realistic that it is difficult for traditional voiceprint recognition systems to distinguish between real human voice and high-quality AI synthesized voice. This poses a risk of being maliciously used for telecommunications fraud and identity theft, and the risk is even higher in smart home scenarios. Summary of the Invention

[0003] Therefore, it is necessary to propose a personalized interaction method and system for smart mirrors that integrates voiceprint recognition to address the problems existing in the current technology.

[0004] The first technical solution of this invention is: A personalized interaction method for a smart mirror integrating voiceprint recognition includes: acquiring current interaction data of a target user and a target smart mirror, wherein the current interaction data includes the target user's current voiceprint feature vector, the target user's speech-text content data, the target user's current environment data, and the target user's current physiological state data, wherein the target user's current voiceprint feature vector and the target user's speech-text content data both correspond to the target user's current original voice commands; acquiring historical interaction data of the target user, wherein the historical interaction data is pre-stored in a target historical interaction database, and the historical interaction data includes the target user's historical normal voiceprint feature vector and the target user's historical normal speech-text content data. This involves analyzing the target user's current interaction data, historical normal environment data, historical physiological state data, and historical normal personalized interaction data. A multi-dimensional similarity comparison is performed between the target user's current interaction data and their historical interaction data to calculate an abnormal interaction index. It is then determined whether this abnormal interaction index is less than a preset threshold. If so, a personalized control command for the target smart mirror is generated using the personalized decision engine of the target smart mirror, based on the target user's current interaction data and their historical normal personalized interaction data. If not, the original voice command is identified as a high-risk abnormal command, and an anti-fraud warning procedure is triggered.

[0005] The second technical solution of this invention is: A personalized interaction system for a smart mirror integrating voiceprint recognition includes: a current data acquisition module, used to acquire current interaction data between a target user and the target smart mirror, wherein the current interaction data includes the target user's current voiceprint feature vector, the target user's speech-text content data, the target user's current environment data, and the target user's current physiological state data, wherein the target user's current voiceprint feature vector and the target user's speech-text content data both correspond to the target user's current original voice commands; and a historical data acquisition module, used to acquire historical interaction data of the target user, wherein the historical interaction data is pre-stored in a target historical interaction database, wherein the historical interaction data includes the target user's historical normal voiceprint feature vector, the target user's historical normal speech-text content data, the target user's historical normal environment data, the target user's historical physiological state data, and the target user's historical normal personal voiceprint feature vector. Personalized interaction data; an anomaly index calculation module, used to perform multi-dimensional similarity comparison between the target user's current interaction data and the target user's historical interaction data, and calculate the target user's current interaction anomaly index; an anomaly index judgment module, used to determine whether the target user's anomaly index is less than the target user's preset anomaly index threshold; a first index processing module, used to generate a target smart mirror personalized control command corresponding to the target user based on the target user's current interaction data and the target user's historical normal personalized interaction data through the target smart mirror's personalized decision engine when the target user's anomaly index is less than the target user's preset anomaly index threshold; a second index processing module, used to determine that the target user's original voice command is a target high-risk abnormal command and trigger an anti-fraud warning program when the target user's anomaly index is greater than or equal to the target user's preset anomaly index threshold.

[0006] The third technical solution of this invention is: A computer device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the following steps: The system acquires the current interaction data between the target user and the target smart mirror. This current interaction data includes the target user's current voiceprint feature vector, the target user's speech-text content data, the target user's current environmental data, and the target user's current physiological state data. The target user's current voiceprint feature vector and the target user's speech-text content data both correspond to the target user's current original voice commands. The system also acquires the target user's historical interaction data, which is pre-stored in a target historical interaction database. This historical interaction data includes the target user's historical normal voiceprint feature vector, the target user's historical normal speech-text content data, and the target user's historical... The system uses normal environmental data, historical physiological state data of the target user, and historical normal personalized interaction data of the target user. It then performs a multi-dimensional similarity comparison between the current interaction data and the historical interaction data of the target user to calculate the target user's abnormal interaction index. It determines whether the target user's abnormal interaction index is less than a preset abnormal interaction index threshold. If so, it generates a personalized control command for the target smart mirror based on the target user's current interaction data and historical normal personalized interaction data, using the personalized decision engine of the target smart mirror. If not, it determines the target user's original voice command as a high-risk abnormal command and triggers an anti-fraud warning program.

[0007] The fourth technical solution of this invention is: A computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps: The system acquires the current interaction data between the target user and the target smart mirror. This current interaction data includes the target user's current voiceprint feature vector, the target user's speech-text content data, the target user's current environmental data, and the target user's current physiological state data. The target user's current voiceprint feature vector and the target user's speech-text content data both correspond to the target user's current original voice commands. The system also acquires the target user's historical interaction data, which is pre-stored in a target historical interaction database. This historical interaction data includes the target user's historical normal voiceprint feature vector, the target user's historical normal speech-text content data, and the target user's historical... The system uses normal environmental data, historical physiological state data of the target user, and historical normal personalized interaction data of the target user. It then performs a multi-dimensional similarity comparison between the current interaction data and the historical interaction data of the target user to calculate the target user's abnormal interaction index. It determines whether the target user's abnormal interaction index is less than a preset abnormal interaction index threshold. If so, it generates a personalized control command for the target smart mirror based on the target user's current interaction data and historical normal personalized interaction data, using the personalized decision engine of the target smart mirror. If not, it determines the target user's original voice command as a high-risk abnormal command and triggers an anti-fraud warning program.

[0008] Compared with the prior art, the present invention has the following beneficial effects: This invention first acquires the current interaction data of the target user and the target smart mirror, then acquires the historical interaction data of the target user, and then performs a multi-dimensional similarity comparison between the current interaction data and the historical interaction data to calculate the abnormal interaction index of the target user's current interaction. Finally, it determines whether the abnormal interaction index is less than a preset abnormal interaction index threshold. If so, it generates personalized control commands for the target smart mirror based on the target user's current interaction data and historical normal personalized interaction data through the personalized decision engine of the target smart mirror. If not, it determines that the target user's original voice command is a high-risk abnormal command and triggers an anti-fraud warning program. This can effectively prevent fraudulent activities such as AI voice imitation and recording attacks, and protect users' property and privacy. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] in: Figure 1 This is a flowchart illustrating one implementation method of the personalized interaction method for a smart mirror that integrates voiceprint recognition in one embodiment. Figure 2 This is a framework diagram of one embodiment of a personalized interactive system for a smart mirror that integrates voiceprint recognition. Figure 3 This is a structural block diagram of one embodiment of a computer device. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] Please see Figure 1 , combined Figure 1 It can be seen that a personalized interaction method for a smart mirror that integrates voiceprint recognition according to an embodiment of the present invention includes the following steps: Step S101: Obtain the target user's current interaction data with the target smart mirror. The target user's current interaction data includes the target user's current voiceprint feature vector, the target user's voice text content data, the target user's current environment data, and the target user's current physiological state data. The target user's current voiceprint feature vector and the target user's voice text content data are both corresponding to the target user's current original voice command.

[0013] Optionally, in this step, obtaining the current interaction data between the target user and the target smart mirror includes: First, obtain the original voice command of the target user corresponding to the target smart mirror, and preprocess the original voice command of the target user to obtain the current voiceprint feature vector of the target user and the voice text content data of the target user corresponding to the original voice command of the target user.

[0014] In this step, the user's original voiceprint signal is collected through the microphone array of the smart mirror. The original voiceprint signal is preprocessed to extract the current voiceprint feature vector. The preprocessing includes noise reduction, frame segmentation, windowing and feature extraction.

[0015] Second, acquire the target user's current environmental data and the target user's current physiological state data. The target user's current environmental data includes the target's geographical location data, the target's real-time weather data, and the target's environmental noise level data. The target user's current physiological state data includes the target user's current heart rate data, the target user's current tone of voice data, and the target user's current speech rate data.

[0016] This step involves acquiring real-time environmental data through environmental sensors and acquiring user physiological status data through a user status monitoring module.

[0017] Third, the target user's current voiceprint feature vector, the target user's voice text content data, the target user's current environment data, and the target user's current physiological state data are used as the target user's current interaction data.

[0018] This step is the starting point of the interaction, transforming the sound signal into computable features and understandable instructions.

[0019] This step incorporates objective factors such as weather and geographical location to provide context for subsequent personalized decision-making and anomaly detection. For example, requesting to open a window during a thunderstorm might be considered anomalous.

[0020] Step S102: Obtain the target user's historical interaction data, wherein the target user's historical interaction data is pre-stored in the target historical interaction database, and the target user's historical interaction data includes the target user's historical normal voiceprint feature vector, the target user's historical normal speech text content data, the target user's historical normal environment data, the target user's historical physiological state data, and the target user's historical normal personalized interaction data.

[0021] Optionally, this step retrieves the user's historical voiceprint feature dataset {HFV_normal, HFV_abnormal} from the historical database, where HFV_normal is the set of normal historical voiceprint features and HFV_abnormal is the set of abnormal historical voiceprint features. This step retrieves not only the user's normal historical voiceprint feature set from the historical database but also the user's abnormal historical voiceprint feature set from the historical database.

[0022] This step is the core of personalization and anti-fraud measures. The system does not judge a single instruction in isolation, but compares it with the target user's long-term normal behavior profile.

[0023] Step S103: Perform a multi-dimensional similarity comparison between the current interaction data of the target user and the historical interaction data of the target user, and calculate the abnormal interaction index of the target user's current interaction.

[0024] Step S104: Determine whether the abnormal interaction index of the target user is less than the preset abnormal interaction index threshold of the target user.

[0025] Step S105: If the abnormal interaction index of the target user is less than the preset abnormal interaction index threshold of the target user, then based on the current interaction data of the target user and the historical normal personalized interaction data of the target user, the personalized decision engine of the target smart mirror generates a personalized control command for the target smart mirror corresponding to the target user.

[0026] This step, after confirming safety, allows the system to enter the service phase, generating the most personalized response based on the environment and the user's historical preferences. For example, when playing music, it might recommend news podcasts during the morning commute and upbeat songs during evening exercise.

[0027] Step S106: If the abnormal interaction index of the target user is greater than or equal to the preset abnormal interaction index threshold of the target user, then the original voice command of the target user is determined to be a high-risk abnormal command and the anti-fraud warning program is triggered.

[0028] In this step, while completing the service, the interaction is stored in the history database as a new normal sample, enabling the system to adapt to the natural changes in the user's voice with age and health status, and achieve continuous evolution.

[0029] Steps S103-S106 are crucial for preventing AI-synthesized speech imitation. By comprehensively calculating the deviation between the user's voice characteristics and environmental parameters through a mathematical model (anomaly detection model), even if the AI ​​imitates the voice, if its command content (such as requesting a money transfer late at night) or the virtual environmental parameters when issuing the command (such as imitating a voice at home but with an IP address overseas) is significantly inconsistent with historical normal patterns, it will be identified as an anomaly and blocked.

[0030] In this embodiment, optionally, before performing a multi-dimensional similarity comparison between the current interaction data of the target user and the historical interaction data of the target user to calculate the abnormal interaction index of the target user's current interaction, the method further includes: The comprehensive similarity score S_total between the current voiceprint feature vector FV_current and the historical dataset is calculated using the following formula: S_total=α×S_normal+β×(1-S_abnormal)+γ×E_factor+δ×G_factor+ε×H_factor; in: S_normal: The similarity between FV_current and HFV_normal. The formula is S_normal=max(cosine_similarity(FV_current,HFV_normal_i)), where i=1,2,...,n.

[0031] S_abnormal: The similarity between FV_current and HFV_abnormal, calculated using the following formula: S_abnormal=max(cosine_similarity(FV_current,HFV_abnormal_j)), j=1,2,...,m.

[0032] E_factor: Environmental correction factor, calculated based on real-time environmental data, E_factor=f(temperature, humidity, noise level, air pressure).

[0033] G_factor: Geographic correction factor, calculated based on the geographic features of the smart mirror's installation location. G_factor = g(altitude, latitude and longitude, building structure).

[0034] H_factor: Health status correction factor, calculated based on user physiological status data, H_factor=h(heart rate variability, speech energy distribution, speech rate stability).

[0035] α,β,γ,δ,ε: weighting coefficients, and α+β+γ+δ+ε=1, dynamically adjusted through machine learning.

[0036] S5: Determine the normality of the current voiceprint data based on the comprehensive similarity score S_total: If S_total≥θ_high, it is determined to be a normal voiceprint, and S6 is executed; If θ_low≤S_total<θ_high, it is determined to be a suspicious voiceprint, and S7 is executed; If S_total < θ_low, it is determined to be an abnormal voiceprint, and S8 is executed; S6: Normal voiceprint processing flow: Confirm user identity, combine environmental data, geographic data and user status data to generate personalized interactive content and display it on the Magic Mirror interface; S7: Suspicious voiceprint processing procedure: Initiate a secondary verification mechanism, including facial recognition, password input or security question verification. If the verification is successful, proceed to S6; if the verification fails, proceed to S8. S8: Abnormal voiceprint processing flow: Record abnormal events, issue security alarms, and implement corresponding protective measures based on the abnormality type.

[0037] In this embodiment, optionally, the step of performing a multi-dimensional similarity comparison between the current interaction data of the target user and the historical interaction data of the target user to calculate the abnormal interaction index of the target user's current interaction includes: The target user's current voiceprint feature vector, the target user's speech-text content data, the target user's current environment data, and the target user's current physiological state data are compared with the target user's historical normal voiceprint feature vector, the target user's historical normal speech-text content data, the target user's historical normal environment data, and the target user's historical physiological state data in a corresponding multidimensional similarity comparison, and the target user's current interaction with the target user's voiceprint similarity abnormality index is calculated.

[0038] In this embodiment, optionally, the target user's abnormal interaction index for the current interaction of the target user is calculated using the following formula: A_score =αxD(V_c,{V_h})+βxD(E_c,{E_h})+γxF(T_c,H_uid) +δxG(P_c,{P_h}); Wherein, A_score is the abnormal interaction index of the target user, D(V_c,{V_h}) is the voiceprint difference degree, V_c is the current voiceprint feature vector of the target user, and V_h is the historical normal voiceprint feature vector of the target user; wherein, D(E_c,{E_h}) is the environmental difference degree, E_c is the current environmental data of the target user, and E_h is the historical normal environmental data of the target user; wherein, F(T_c,H_uid) is the behavioral abnormality degree, T_c is the voice-text content data of the target user, and H_uid is the historical normal voice-text content data of the target user; wherein, G(P_c,{P_h}) is the user physiological state difference degree, P_c is the current physiological state data of the target user, and P_h is the historical physiological state data of the target user; wherein, α, β, γ, and δ are all weighting coefficients, and α+β+γ+δ=1.

[0039] In this embodiment, optionally, after generating the target smart mirror personalized control command corresponding to the target user through the personalized decision engine of the target smart mirror based on the target user's current interaction data and the target user's historical normal personalized interaction data, the process includes: First, the personalized control commands of the target smart mirror are executed through the target display and broadcasting module corresponding to the target smart mirror.

[0040] Second, after executing the personalized control command of the target smart mirror, the current interaction data and interaction result of the target user in this interaction are updated to the target historical interaction database.

[0041] In this embodiment, optionally, after determining that the target user's original voice command is a high-risk abnormal command and triggering the anti-fraud warning procedure, the following steps are included: First, a secondary verification mechanism is activated to obtain the secondary voice command of the target user corresponding to the target user, and to determine whether the secondary voice command of the target user is a live voice.

[0042] Second, if the target user's secondary voice command is not a live voice, then the target user's secondary voice command is determined to be a high-risk abnormal command, and the anti-fraud warning procedure is triggered.

[0043] Third, if the target user's secondary voice command is a live voice, then the target user is subjected to facial recognition, password input, or security question verification, and it is determined whether the target user can pass the verification.

[0044] Fourth, if so, then based on the target user's current interaction data and the target user's historical normal personalized interaction data, the target smart mirror's personalized decision engine generates the target smart mirror's personalized control command corresponding to the target user.

[0045] Fifth, if not, the target user's secondary voice command is determined to be a high-risk abnormal command, and the anti-fraud warning procedure is triggered.

[0046] In this embodiment, optionally, after obtaining the current interaction data between the target user and the target smart mirror, the process includes: First, the target user's original voice commands are sequentially subjected to spectral discontinuity detection, phase consistency detection, breath sound detection, and digital trace detection to obtain the confidence level of the target user's synthesized voice.

[0047] Second, determine whether the confidence level of the synthesized speech of the target user is greater than or equal to the target preset confidence level threshold for synthesized speech.

[0048] Third, if the confidence level of the target user's synthesized speech is greater than or equal to the target preset synthesized speech confidence level threshold, then the target user's original voice command is determined to be AI synthesized speech, and the target user's original voice command is determined to be the target high-risk abnormal command, triggering the anti-fraud warning procedure.

[0049] Specifically, the original voice commands of the target user are sequentially subjected to spectral discontinuity detection, phase consistency detection, breath sound detection, and digit trace detection to obtain the confidence level of the synthesized voice of the target user, including: Synthetic speech feature detection is performed on the preprocessed speakerprint signal, and the synthesized speech confidence score (Score_synthetic) of the target user is calculated. Score_synthetic=ω1×C_spectral+ω2×C_phase+ω3×C_breath+ω4×C_artifacts; in: C_spectral: Spectral discontinuity score, which detects abrupt changes and unnatural smoothness in the speech spectrum in the frequency domain.

[0050] C_phase: Phase consistency score, which detects unnatural distributions of phase information.

[0051] C_breath: Breath sound abnormality score, which detects unnatural absence or patterns of breath sounds in synthesized speech.

[0052] C_artifacts: Synthetic artifact scores, which detect unique digital artifacts in AI synthesis using a pre-trained deep neural network. C_artifacts=DNN_synthetic_detector(FV_current); ω1, ω2, ω3, ω4: Weight coefficients, obtained through adversarial example training.

[0053] If Score_synthetic ≥ θ_synthetic, it is determined to be AI synthesized speech, and the highest level of anomaly handling and security alert is executed directly.

[0054] The significance of spectral discontinuity detection lies in the fact that AI-generated sounds are often too perfect on the spectrogram, lacking the subtle fluctuations and discontinuities of real human voices. This indicator is used to calculate the "naturalness" of spectral changes.

[0055] The significance of phase consistency detection lies in the fact that the phase of real sound waves has a specific natural pattern, while AI-synthesized speech will have subtle anomalies in phase distribution, just like distinguishing the brushstrokes of a real painting from a fake one.

[0056] The significance of breath sound detection lies in the fact that real people have natural breathing sounds when they speak, while AI-generated speech often lacks these breathing sounds or has an unnatural breathing pattern. For example, real people inhale before speaking long sentences and exhale after speaking, while AI often ignores this detail.

[0057] The significance of digital trace detection lies in using deep learning networks to directly analyze whether speech contains digital fingerprints unique to AI synthesis tools, just like checking the EXIF ​​information of a photo to determine whether it has been photoshopped.

[0058] In short, the four indicators are weighted and summed. If the total score exceeds the threshold, it is determined to be an AI-synthesized voice, and an alarm is immediately triggered.

[0059] Please see Figure 2 , combined Figure 2 It can be seen that the intelligent magic mirror personalized interaction system 100 integrating voiceprint recognition according to an embodiment of the present invention includes: The current data acquisition module 10 is used to acquire the current interaction data of the target user and the target smart mirror. The current interaction data of the target user includes the current voiceprint feature vector of the target user, the voice text content data of the target user, the current environment data of the target user and the current physiological state data of the target user. The current voiceprint feature vector of the target user and the voice text content data of the target user are both corresponding to the target user's current original voice command. The historical data acquisition module 20 is used to acquire the target user's historical interaction data corresponding to the target user. The target user's historical interaction data is pre-stored in the target historical interaction database. The target user's historical interaction data includes the target user's historical normal voiceprint feature vector, the target user's historical normal speech text content data, the target user's historical normal environment data, the target user's historical physiological state data, and the target user's historical normal personalized interaction data. The abnormality index calculation module 30 is used to perform a multi-dimensional similarity comparison between the current interaction data of the target user and the historical interaction data of the target user, and calculate the abnormal interaction index of the target user's current interaction. The abnormal index judgment module 40 is used to determine whether the abnormal interaction index of the target user is less than the preset abnormal interaction index threshold of the target user. The first index processing module 50 is used to generate a personalized control command for the target smart mirror corresponding to the target user based on the target user's current interaction data and the target user's historical normal personalized interaction data when the target user's abnormal interaction index is less than the target user's preset abnormal interaction index threshold. The second index processing module 60 is used to determine that the target user's original voice command is a high-risk abnormal command and trigger an anti-fraud warning program when the target user's abnormal interaction index is greater than or equal to the target user's preset abnormal interaction index threshold.

[0060] Figure 3An internal structural diagram of a computer device in one embodiment is shown. This computer device can specifically be a terminal or a server. Figure 3 As shown, the computer device includes a processor, a memory, and a network interface connected via a device bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores operating devices and may also store a computer program. When executed by the processor, this computer program enables the processor to implement the aforementioned personalized interaction method for the smart mirror that integrates voiceprint recognition. The internal memory may also store a computer program, which, when executed by the processor, enables the processor to execute the aforementioned personalized interaction method for the smart mirror that integrates voiceprint recognition. Those skilled in the art will understand that… Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0061] In another embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the following steps: The system acquires the current interaction data between the target user and the target smart mirror. This current interaction data includes the target user's current voiceprint feature vector, the target user's speech-text content data, the target user's current environmental data, and the target user's current physiological state data. The target user's current voiceprint feature vector and the target user's speech-text content data both correspond to the target user's current original voice commands. The system also acquires the target user's historical interaction data, which is pre-stored in a target historical interaction database. This historical interaction data includes the target user's historical normal voiceprint feature vector, the target user's historical normal speech-text content data, and the target user's historical... The system uses normal environmental data, historical physiological state data of the target user, and historical normal personalized interaction data of the target user. It then performs a multi-dimensional similarity comparison between the current interaction data and the historical interaction data of the target user to calculate the target user's abnormal interaction index. It determines whether the target user's abnormal interaction index is less than a preset abnormal interaction index threshold. If so, it generates a personalized control command for the target smart mirror based on the target user's current interaction data and historical normal personalized interaction data, using the personalized decision engine of the target smart mirror. If not, it determines the target user's original voice command as a high-risk abnormal command and triggers an anti-fraud warning program.

[0062] In another embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, causes the processor to perform the following steps: The system acquires the current interaction data between the target user and the target smart mirror. This current interaction data includes the target user's current voiceprint feature vector, the target user's speech-text content data, the target user's current environmental data, and the target user's current physiological state data. The target user's current voiceprint feature vector and the target user's speech-text content data both correspond to the target user's current original voice commands. The system also acquires the target user's historical interaction data, which is pre-stored in a target historical interaction database. This historical interaction data includes the target user's historical normal voiceprint feature vector, the target user's historical normal speech-text content data, and the target user's historical... The system uses normal environmental data, historical physiological state data of the target user, and historical normal personalized interaction data of the target user. It then performs a multi-dimensional similarity comparison between the current interaction data and the historical interaction data of the target user to calculate the target user's abnormal interaction index. It determines whether the target user's abnormal interaction index is less than a preset abnormal interaction index threshold. If so, it generates a personalized control command for the target smart mirror based on the target user's current interaction data and historical normal personalized interaction data, using the personalized decision engine of the target smart mirror. If not, it determines the target user's original voice command as a high-risk abnormal command and triggers an anti-fraud warning program.

[0063] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0064] This invention first acquires the current interaction data of the target user and the target smart mirror, and then acquires the historical interaction data of the target user. Next, it performs a multi-dimensional similarity comparison between the current and historical interaction data to calculate the abnormal interaction index of the target user's current interaction. Finally, it determines whether the abnormal interaction index is less than a preset threshold. If so, it generates a personalized control command for the target smart mirror based on the target user's current and historical normal personalized interaction data using the smart mirror's personalized decision engine. If not, it determines the target user's original voice command as a high-risk abnormal command and triggers an anti-fraud warning program. The beneficial effects of this invention are as follows: First, it offers high security. Through multi-dimensional anomaly detection and dynamic challenge response, it effectively prevents fraudulent activities such as AI voice imitation and recording attacks, protecting users' property and privacy.

[0065] Second, it is highly personalized, combining the environment (weather, location) and users' historical habits to provide interactive content and services that are tailored to each individual and the time and place.

[0066] Third, it is a continuous evolution. By constantly learning new normal user interaction data, the system can adapt to natural changes in voice and avoid false rejections due to reasons such as users having a cold or voice changes.

[0067] Fourth, it is environmentally adaptive. By sensing parameters such as environmental noise, it automatically adjusts the audio processing algorithm to improve the voiceprint recognition rate in different scenarios.

[0068] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0069] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A personalized interaction method for a smart mirror integrating voiceprint recognition, characterized in that, include: The target user's current interaction data with the target smart mirror is obtained. The target user's current interaction data includes the target user's current voiceprint feature vector, the target user's voice text content data, the target user's current environment data, and the target user's current physiological state data. The target user's current voiceprint feature vector and the target user's voice text content data are both corresponding to the target user's current original voice command. Obtain the target user's historical interaction data corresponding to the target user, wherein the target user's historical interaction data is pre-stored in the target historical interaction database, and the target user's historical interaction data includes the target user's historical normal voiceprint feature vector, the target user's historical normal speech text content data, the target user's historical normal environment data, the target user's historical physiological state data, and the target user's historical normal personalized interaction data. The current interaction data of the target user and the historical interaction data of the target user are compared in multiple dimensions to calculate the abnormal interaction index of the target user in the current interaction. Determine whether the abnormal interaction index of the target user is less than the preset abnormal interaction index threshold of the target user; If so, then based on the target user's current interaction data and the target user's historical normal personalized interaction data, the target smart mirror's personalized control command corresponding to the target user is generated through the target smart mirror's personalized decision engine; If not, the original voice command of the target user is determined to be a high-risk abnormal command, and the anti-fraud warning procedure is triggered.

2. The personalized interaction method for a smart mirror integrating voiceprint recognition according to claim 1, characterized in that, The step of obtaining the current interaction data between the target user and the target smart mirror includes: Obtain the original voice command of the target user corresponding to the target smart mirror, and preprocess the original voice command of the target user to obtain the current voiceprint feature vector of the target user and the voice text content data of the target user corresponding to the original voice command of the target user; The target user's current environmental data and target user's current physiological state data are obtained. The target user's current environmental data includes target geographical location data, target real-time weather data, and target environmental noise level data. The target user's current physiological state data includes target user's current heart rate data, target user's current tone of voice data, and target user's current speech rate data. The target user's current voiceprint feature vector, the target user's voice text content data, the target user's current environment data, and the target user's current physiological state data are used as the target user's current interaction data.

3. The personalized interaction method for a smart mirror integrating voiceprint recognition according to claim 1, characterized in that, The step of performing a multi-dimensional similarity comparison between the target user's current interaction data and the target user's historical interaction data to calculate the target user's abnormal interaction index for the current interaction includes: The target user's current voiceprint feature vector, the target user's speech-text content data, the target user's current environment data, and the target user's current physiological state data are compared with the target user's historical normal voiceprint feature vector, the target user's historical normal speech-text content data, the target user's historical normal environment data, and the target user's historical physiological state data in a corresponding multidimensional similarity comparison, and the target user's current interaction with the target user's voiceprint similarity abnormality index is calculated.

4. The personalized interaction method for a smart mirror integrating voiceprint recognition according to claim 1, characterized in that, The target user's abnormal interaction index is calculated using the following formula: A_score =αxD(V_c,{V_h})+βxD(E_c,{E_h})+γxF(T_c,H_uid) +δxG(P_c,{P_h}); Wherein, A_score is the abnormal interaction index of the target user, D(V_c,{V_h}) is the voiceprint difference degree, V_c is the current voiceprint feature vector of the target user, and V_h is the historical normal voiceprint feature vector of the target user; wherein, D(E_c,{E_h}) is the environmental difference degree, E_c is the current environmental data of the target user, and E_h is the historical normal environmental data of the target user; wherein, F(T_c,H_uid) is the behavioral abnormality degree, T_c is the voice-text content data of the target user, and H_uid is the historical normal voice-text content data of the target user; wherein, G(P_c,{P_h}) is the user physiological state difference degree, P_c is the current physiological state data of the target user, and P_h is the historical physiological state data of the target user; wherein, α, β, γ, and δ are all weighting coefficients, and α+β+γ+δ=1.

5. The personalized interaction method for a smart mirror integrating voiceprint recognition according to claim 1, characterized in that, After generating personalized control instructions for the target smart mirror corresponding to the target user through the personalized decision engine of the target smart mirror based on the target user's current interaction data and the target user's historical normal personalized interaction data, the process includes: The personalized control commands of the target smart mirror are executed through the target display and broadcasting module corresponding to the target smart mirror. After executing the personalized control command of the target smart mirror, the current interaction data and interaction result of the target user in this interaction are updated to the target historical interaction database.

6. The personalized interaction method for a smart mirror integrating voiceprint recognition according to claim 1, characterized in that, After determining that the target user's original voice command is a high-risk abnormal command and triggering the anti-fraud warning procedure, the procedure includes: A secondary verification mechanism is enabled to obtain the secondary voice command of the target user corresponding to the target user, and to determine whether the secondary voice command of the target user is a live voice. If not, the target user's secondary voice command is determined to be a high-risk abnormal command, and the anti-fraud warning procedure is triggered. If so, then perform facial recognition, password input, or security question verification on the target user, and determine whether the target user can pass the verification; If so, then based on the target user's current interaction data and the target user's historical normal personalized interaction data, the target smart mirror's personalized decision engine generates the target smart mirror's personalized control command corresponding to the target user; If not, the target user's secondary voice command is determined to be a high-risk abnormal command, and the anti-fraud warning procedure is triggered.

7. The personalized interaction method for a smart mirror integrating voiceprint recognition according to claim 1, characterized in that, After obtaining the current interaction data between the target user and the target smart mirror, the process includes: The original voice commands of the target user are sequentially subjected to spectral discontinuity detection, phase consistency detection, breath sound detection, and digital trace detection to obtain the confidence level of the synthesized voice of the target user; Determine whether the confidence level of the synthesized speech of the target user is greater than or equal to the target preset confidence threshold for synthesized speech; If so, the original voice command of the target user is determined to be AI-synthesized voice, and the original voice command of the target user is determined to be a high-risk abnormal command of the target, triggering the anti-fraud warning program.

8. A personalized interactive system for a smart mirror integrating voiceprint recognition, characterized in that, include: The current data acquisition module is used to acquire the current interaction data of the target user and the target smart mirror. The current interaction data of the target user includes the current voiceprint feature vector of the target user, the voice text content data of the target user, the current environment data of the target user, and the current physiological state data of the target user. The current voiceprint feature vector of the target user and the voice text content data of the target user are both corresponding to the target user's current original voice command. The historical data acquisition module is used to acquire the target user's historical interaction data, wherein the target user's historical interaction data is pre-stored in the target historical interaction database, and the target user's historical interaction data includes the target user's historical normal voiceprint feature vector, the target user's historical normal speech text content data, the target user's historical normal environmental data, the target user's historical physiological state data, and the target user's historical normal personalized interaction data. The abnormality index calculation module is used to perform a multi-dimensional similarity comparison between the current interaction data of the target user and the historical interaction data of the target user, and calculate the abnormal interaction index of the target user's current interaction. An abnormal index judgment module is used to determine whether the abnormal interaction index of the target user is less than the preset abnormal interaction index threshold of the target user. The first index processing module is used to generate a personalized control command for the target smart mirror corresponding to the target user based on the target user's current interaction data and the target user's historical normal personalized interaction data when the target user's abnormal interaction index is less than the target user's preset abnormal interaction index threshold. The second index processing module is used to determine that the target user's original voice command is a high-risk abnormal command and trigger an anti-fraud warning program when the target user's abnormal interaction index is greater than or equal to the target user's preset abnormal interaction index threshold.

9. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, causes the processor to perform the personalized interaction method for a smart mirror that integrates voiceprint recognition as described in any one of claims 1-7.

10. A computer device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor causes the processor to perform the intelligent smart mirror personalized interaction method that integrates voiceprint recognition as described in any one of claims 1-7.

Images