Emotion Perception and Guest Room Management Method Based on Voiceprint Recognition

Abstract

This application provides a voiceprint recognition-based emotion perception and room management method, comprising: generating a service configuration adjustment plan that matches the user's emotional tendencies based on demand prediction results; fine-tuning the service content based on real-time changes captured by dynamic data to obtain the adjusted service parameters; if the optimized emotional cycle change pattern deviates from the actual user feedback, then updating the voiceprint feature matrix and emotional tendency distribution by collecting long-term data records again to determine the new periodic fluctuation characteristics; and based on the new periodic fluctuation characteristics, cyclically executing the construction of the emotional profile model and the personalized demand prediction process to continuously improve the targeting of service configuration adjustments and obtain the final optimized service quality improvement plan.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for emotion perception and room management based on voiceprint recognition. Background Technology

[0002] Voiceprint recognition technology, as an important branch of artificial intelligence and affective computing, has undeniable value in enhancing personalized services and user experience, especially demonstrating great potential in hotel room management. By analyzing human voice characteristics to perceive emotional states, it can not only help the service industry more accurately understand customer needs but also provide data support for optimizing service processes. However, current methods still have significant shortcomings, primarily in their lack of ability to dynamically capture and analyze user emotional data over long periods. Many solutions tend to focus only on single scenarios or short-term voice data collection, ignoring the complex patterns of user emotions changing over different times and environments, resulting in a lack of continuity and targeted adjustments in service.

[0003] Against this backdrop, the core challenges facing this field are becoming increasingly apparent. First, the cyclical patterns of emotional state changes are difficult to identify effectively. Because emotions are influenced by multiple factors such as time and context, relying solely on fragmented data records cannot construct a complete user emotional profile, thus limiting the accuracy of service predictions. Second, this insufficient identification of cyclical changes further complicates the prediction of personalized needs. The failure to extract users' emotional tendencies over specific time periods from long-term data means that service configurations cannot truly meet user expectations and may even lead to user dissatisfaction. These two challenges are progressively more complex and together constitute the bottleneck of current technology applications.

[0004] Therefore, how to identify the periodic changes in users' voiceprint emotion data through long-term recording and analysis, and accurately predict their emotional needs in specific time periods, has become a key issue in improving the quality of guest room management services. Summary of the Invention

[0005] This invention provides a method for emotion perception and guest room management based on voiceprint recognition, mainly including:

[0006] The voice data of users in the room management scenario is recorded in real time by voiceprint acquisition device. The voice signals are initially classified according to different time periods and environmental conditions to obtain the raw voiceprint dataset for subsequent emotion state analysis and feature extraction, resulting in a set of voiceprint signals after preliminary classification.

[0007] For the set of voiceprint signals after preliminary classification, voiceprint feature extraction technology is used to perform in-depth processing on each sound signal to separate the acoustic feature parameters related to emotions, construct a voiceprint feature matrix containing multi-dimensional features such as pitch, volume and speech rate, and determine the basic data of emotion features.

[0008] Based on the voiceprint feature matrix, time series analysis is used to mine the patterns of long-term recorded emotional feature data, identify the emotional cycle changes of users in different time periods, and determine the time nodes and trends of periodic fluctuations.

[0009] If the identified periodic fluctuation time points and trends have statistically significant regularities, then a user emotion profile model is constructed based on these regularities, combining long-term data records with scene influencing factors to obtain the distribution of users' emotional tendencies in a specific time period.

[0010] By using a user sentiment profiling model, we can dynamically analyze the distribution of sentiment tendencies over a specific time period. By combining time period divisions and scenario influencing factors, we can predict users' personalized needs in future time periods and determine the demand prediction results.

[0011] Based on the demand forecast results, a service configuration adjustment plan matching the user's emotional tendencies is generated. In response to real-time changes captured by dynamic data, the service content is fine-tuned, and the adjusted service parameters are obtained.

[0012] By adjusting the service parameters, the service process of the guest room management system is updated in real time. Combined with the feedback data of service configuration adjustment, the user emotional profile model is continuously optimized to obtain the optimized emotional cycle change pattern.

[0013] If the optimized emotional cycle change pattern deviates from the actual user feedback, then by collecting long-term data records again, updating the voiceprint feature matrix and emotional tendency distribution, new periodic fluctuation characteristics can be identified.

[0014] Based on the new cyclical fluctuation characteristics, the process of building the sentiment profile model and predicting personalized needs is executed cyclically to continuously improve the targeting of service configuration adjustments and obtain the final optimized service quality improvement solution.

[0015] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0016] This invention discloses a method for emotion perception and guest room management based on voiceprint recognition. By collecting and classifying user voice data in real time, extracting a voiceprint feature matrix, analyzing the cyclical changes in emotions, and constructing a user emotion profile model, the system can predict users' personalized needs during specific time periods and dynamically adjust service configurations accordingly, achieving intelligent and personalized guest room management services. This invention, through voiceprint collection, feature extraction, and time series analysis, achieves accurate perception and prediction of users' emotional states, proactively adjusting service content to meet user needs, effectively improving the quality of guest room management services and user satisfaction. Furthermore, this invention has a self-optimization function, continuously improving model accuracy based on actual feedback, thus continuously enhancing service quality. Attached Figure Description

[0017] Figure 1 This is a flowchart of an emotion perception and guest room management method based on voiceprint recognition according to the present invention.

[0018] Figure 2 This is a schematic diagram of an emotion perception and room management method based on voiceprint recognition according to the present invention. Figure 1 .

[0019] Figure 3 This is a schematic diagram of an emotion perception and room management method based on voiceprint recognition according to the present invention. Figure 2 . Detailed Implementation

[0020] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0021] like Figure 1-3 This embodiment discloses a method for emotion perception and room management based on voiceprint recognition, including:

[0022] Step S101: Record the user's voice data in the room management scenario in real time using a voiceprint acquisition device, perform preliminary classification of the voice signals under different time periods and environmental conditions, obtain the original voiceprint dataset, and use it for subsequent emotion state analysis and feature extraction to obtain a set of preliminarily classified voiceprint signals.

[0023] User voice signals are recorded in a guest room management scenario using a voiceprint acquisition device, and a first voiceprint dataset is obtained using microphone array technology. Based on the first voiceprint dataset, environmental noise is processed using a filter to obtain a second voiceprint dataset. If the signal-to-noise ratio of the second voiceprint dataset is higher than a preset threshold, it is classified according to time period and environmental conditions using a classification algorithm to obtain a first classification signal set. Frequency and amplitude features are extracted from the first classification signal set, and the first classification signal set is processed using Fast Fourier Transform (FFT) technology to obtain a first feature set, which is stored in a database. The first feature set is matched against a preset emotional state template to determine the corresponding emotional category. Based on the emotional category and the first feature set, a classification report is generated and stored in the database, determining the final voiceprint signal analysis result.

[0024] Specifically, in the guest room management scenario, user voice data is recorded in real time using a voiceprint acquisition device. A high-sensitivity microphone array can be used, with a sampling rate of 44.1kHz and a bit depth of 16bit to ensure high-quality sound signal capture, covering voice data from different locations within the guest room. Next, the sound signals are preliminarily classified according to different time periods and environmental conditions. A time-domain analysis algorithm divides the day into three peak periods: 6:00-9:00 AM, 11:00-2:00 PM, and 6:00-10:00 PM. Combined with an environmental noise detection module, background noise intensity is identified (e.g., below 30dB indicates a quiet environment, above 50dB indicates a noisy environment). Short-time energy analysis and zero-crossing rate calculation are then used to preliminarily separate the sound signals into speech and non-speech signals, resulting in the original voiceprint dataset, where speech signals account for approximately 60%. Subsequently, for further emotion state analysis and feature extraction, Mel-frequency cepstral coefficients (MFCC) algorithm is used to extract voiceprint features. With a frame length of 25ms and a frame shift of 10ms, 13th-order MFCC coefficients are calculated, and principal component analysis (PCA) is used for dimensionality reduction, retaining 95% of the feature variance to ensure feature effectiveness. In this process, for the initially classified voiceprint signal set, a support vector machine (SVM) classifier is used to further subdivide the signal categories. The kernel function is set to radial basis function (RBF), and the parameter C is optimized to 1.0, achieving a classification accuracy of over 85%, forming a classified voiceprint signal set. These steps are closely linked through algorithms and data processing; the classification result of the previous step directly affects the feature extraction accuracy of the next step. Furthermore, by integrating with the room management scenario, such as adjusting the classification threshold based on time period and environmental noise, the speech recognition rate in noisy environments is ensured to be no less than 80%, thus providing reliable data support for subsequent emotion analysis.

[0025] Step S102: For the set of voiceprint signals after preliminary classification, voiceprint feature extraction technology is used to perform in-depth processing on each sound signal to separate the acoustic feature parameters related to emotions, construct a voiceprint feature matrix containing multi-dimensional features such as pitch, volume and speech rate, and determine the basic data of emotion features.

[0026] Initial data is obtained from the voiceprint signal. A feature extraction tool is used to separate emotion-related acoustic features, resulting in a preliminary acoustic feature set. Based on this preliminary acoustic feature set, pitch parameters, volume variations, and speech rate rhythm are quantized to determine corresponding multi-dimensional feature data. Using this multi-dimensional feature data, a feature matrix containing pitch parameters, volume variations, and speech rate rhythm is constructed, resulting in a comprehensive feature matrix structure. If the feature value of a certain dimension in the comprehensive feature matrix structure deviates from a preset threshold range, the dimensional data is standardized to obtain an adjusted feature matrix. Based on the adjusted feature matrix, a support vector machine algorithm is used to classify the emotion data and determine the distribution of emotion-related features. For the classified emotion data, a mapping relationship corresponding to the basic data is obtained to determine the final emotion feature base dataset. The final emotion feature base dataset is then validated using a pre-established acoustic feature model to obtain validated emotion feature data.

[0027] Specifically, for the initially classified voiceprint signal set, each audio signal is first preprocessed using digital signal processing techniques. Assuming the input audio signal sampling rate is 16kHz, the signal is framed using a Hanning window, with each frame being 25 milliseconds long and a frame shift of 10 milliseconds to reduce spectral leakage. Then, the spectral features of each frame are calculated using Short-Time Fourier Transform (STFT), extracting the fundamental frequency (F0) as the pitch feature, calculated within the range of 75Hz to 300Hz. This is combined with the first 13 coefficients of the Mel-frequency cepstral coefficients (MFCC) to capture timbre variations, forming a preliminary feature set. Next, for volume features, the short-time energy of each frame is calculated, and a logarithmic transform is used to normalize the energy value to between 0 and 1. For example, an energy value of 0.85 for a certain frame reflects sound intensity. Speech rate features are quantified by detecting the ratio of pause time to effective speech time. Assuming a 3-second audio clip has an effective speech time of 2.1 seconds, the speech rate ratio is 0.7, reflecting the speaking rhythm. Furthermore, the extracted multi-dimensional features, such as pitch, volume, and speech rate, are integrated into a voiceprint feature matrix. Assuming the matrix has an N×15 dimension, where N is the number of frames and 15 represents the feature dimensions (including F0, 13 MFCC coefficients, and short-time energy), principal component analysis (PCA) is performed on the matrix to reduce the dimensionality to the dimensions with a major contribution rate of 95%, such as reducing it to 10 dimensions to reduce redundant information. Based on this, statistical analysis of emotion-related acoustic features is combined to calculate the mean and variance of each feature dimension; for example, the pitch mean is 120Hz and the variance is 25Hz², serving as the basic data for emotion features. This data can be used for subsequent emotion classification model training, forming a complete logical chain from signal processing to feature construction, ensuring that the feature matrix effectively represents emotional fluctuations.

[0028] Step S103: Based on the voiceprint feature matrix, use time series analysis to mine the patterns of long-term recorded emotional feature data, identify the emotional cycle change patterns of users in different time periods, and determine the time nodes and trends of periodic fluctuations.

[0029] Voiceprint feature data is acquired to construct an initial feature matrix, which contains long-term recorded voiceprint feature data. The initial feature matrix is ​​normalized using a standardization method to obtain a processed feature matrix. Based on the processed feature matrix, time series analysis is applied to decompose the feature distribution across different time periods, yielding a fluctuation sequence of emotional features. Using this fluctuation sequence and a preset threshold range, the periodic variation pattern of emotional features is determined. If the fluctuation sequence exceeds the threshold range within a certain time period, that time period is marked as a significant fluctuation interval, resulting in a set of significant fluctuation time nodes. Based on this set of significant fluctuation time nodes, a sliding window method is used to analyze the time interval and fluctuation amplitude between adjacent nodes, extracting local change trends to obtain the direction of emotional change within each time period. Using these local change trends and the historical distribution of the long-term recorded data, an emotional cycle fluctuation model is constructed to determine the main time nodes of the periodic fluctuations. Based on these main time nodes, K-means clustering is used to classify the emotional feature data, obtaining the category distribution of emotional changes within different time periods. Using this category distribution and the continuity characteristics of the fluctuation sequence, the transition patterns between categories are analyzed to obtain the trend of the user's emotional cycle changes.

[0030] Specifically, for emotional feature data mining based on the voiceprint feature matrix, the first step is to preprocess the long-term recorded voiceprint data. Assuming we collected daily voiceprint data from users for six consecutive months, daily feature values ​​were extracted, including the mean pitch, standard deviation of volume, and rate of change of speech rate, forming a feature matrix containing 540 time points (6 months × 30 days × 3 times / day). Next, time series analysis methods were used to mine patterns in this data. An Autoregressive Moving Average (ARIMA) model was used for modeling, with parameters set to (p=1, d=1, q=1). By differentiating the pitch mean sequence to eliminate trend fluctuations, the autocorrelation coefficient of the residual sequence was calculated, revealing a significant period of 7 days, indicating a possible weekly cycle in user emotions. Subsequently, the time nodes of periodic fluctuations were identified. Fourier transform was used to perform frequency domain analysis on the feature matrix, extracting a period of 7.2 days corresponding to the dominant frequency. Combined with timestamp data, it was determined that Monday morning at 9 AM is the emotional trough and Friday afternoon at 5 PM is the emotional peak, with an error range controlled within ±2 hours. Further analysis of the changing trends revealed that the slope of the average pitch value was calculated using a sliding window method (14-day window). The slope was found to be positive (0.15) in the first week of each month, indicating an upward trend in mood, while the slope was negative (0.12) in the fourth week, indicating a downward trend. Finally, by integrating this periodic pattern with business operations, a mood fluctuation report was automatically generated. Based on the identified 7-day cycle and key milestones, the system predicts the timing of mood lows for the coming week and integrates this with the user's schedule. If a low point coincides with an important meeting, the system automatically adjusts the meeting reminder frequency from the default 2 hours to 4 hours, ensuring users are adequately prepared and forming a data-driven intelligent decision-making closed loop.

[0031] Step S104: If the identified periodic fluctuation time nodes and change trends have statistically significant regularities, then construct a user emotion profile model based on these regularities, combining long-term data records with scene influencing factors to obtain the distribution of user emotion tendencies in a specific time period.

[0032] An initial dataset is constructed by extracting relevant data on periodic fluctuations and time points from historical records. Noise in the initial dataset is removed using preprocessing techniques, resulting in a cleaned fluctuation dataset. Based on this cleaned dataset, the relationship between time points and specific times is analyzed to determine if significant periodic patterns exist. If the fluctuation data exhibits a recurring pattern within a specific time period, it is identified as periodic data. For this periodic data, a feature matrix is ​​constructed, and the correlation between influencing factors and changing trends is integrated to obtain a characteristic emotion-related dataset. Using this characteristic emotion-related dataset, a support vector machine algorithm is applied to train a user model, deriving the classification results of emotion tendencies and obtaining a preliminary emotion tendency distribution. Based on this preliminary emotion tendency distribution, combined with data from specific times and the periodic fluctuations, the distribution results are corrected in the time dimension. If the corrected distribution deviation exceeds a preset threshold, the feature matrix is ​​readjusted to obtain corrected emotion distribution data. This corrected emotion distribution data is mapped onto the user model to construct a complete emotion profile. It is then determined whether the emotion profile covers all specific time periods, resulting in the final user emotion profile. Based on the final user emotion profile results, the distribution derivation and change trend are correlated to generate emotion tendency prediction data for a specific time period, thereby completing the dynamic monitoring of user emotions.

[0033] Specifically, an initial dataset is constructed by extracting relevant data on periodic fluctuations and time points from historical records. Noise is removed through preprocessing to obtain a cleaned fluctuation dataset. Based on this cleaned dataset, the relationship between time points and specific times is analyzed to determine if significant periodic patterns exist. If the fluctuation data exhibits a recurring pattern within a specific time period, it is identified as periodic data. For the identified periodic data, a feature matrix is ​​constructed by combining long-term data and contextual factors. The correlation between influencing factors and changing trends is integrated to obtain a characteristic emotion-related dataset. Using this characteristic emotion-related dataset, a support vector machine algorithm is applied to train the user model, deriving the classification results of emotion tendencies and obtaining a preliminary emotion tendency distribution. Based on this preliminary emotion tendency distribution, and combined with data on specific times and periodic fluctuations, the distribution results are corrected in the time dimension. If the corrected distribution deviation exceeds a preset threshold, the feature matrix is ​​readjusted to obtain corrected emotion distribution data. This corrected emotion distribution data is mapped onto the user model to construct a complete emotion profile. The accuracy of the emotion profile is then determined to cover all specific time periods, resulting in the final user emotion profile. Based on the final user sentiment profile results, the correlation distribution derivation and change trend are used to generate sentiment tendency prediction data for a specific time period, thus completing the dynamic monitoring of user sentiment.

[0034] By analyzing user social media data, we identified periodic fluctuation time points and trends. Assuming that daily text data posted by users was collected from January 2024 to March 2025, including emotional keywords (e.g., "happy" appeared 500 times, "anxious" appeared 300 times), we first used the Fast Fourier Transform (FFT) algorithm to perform time-series analysis on the frequency of emotional keywords, calculating periodic fluctuations. We found that the keyword "happy" had peak frequencies on Mondays and Fridays (an average of 50 times on Mondays and 45 times on Fridays), while "anxious" had the highest frequency on Wednesdays (an average of 35 times), with a cycle of 7 days. The significance test p-value < 0.05 confirmed the existence of the pattern. Next, we constructed a user emotion profile model, using the K-means clustering algorithm (K=3) to classify emotions into three categories: positive, neutral, and negative. Based on word frequency vectors (1000 dimensions, including keywords and contextual features), the clustering results showed that 60% of the data were positive emotions, 25% were neutral, and 15% were negative, with Euclidean distances from the centroids of 0.8, 0.6, and 0.9, respectively. Combining long-term data with contextual factors, user location (70% in city A), weather (60% sunny), and holidays (a 20% increase in positive emotions during the Spring Festival) were used as feature inputs. A random forest algorithm was employed to predict mood tendencies, achieving 85% accuracy on the training set and 80% on the test set. Finally, the mood distribution for a specific time period (e.g., Monday morning 9:00-11:00) was calculated, yielding a 65% probability of positive emotions, 20% of neutral emotions, and 15% of negative emotions. A visual heatmap was used to illustrate the changing trends of emotions over time and context, ensuring consistency between the model's predictions and actual scenarios (e.g., holiday effects), thus forming a closed-loop validation.

[0035] Step S105: Through the user emotion profile model, the distribution of emotion tendencies in a specific time period is dynamically analyzed. Combining the time period division and scene influencing factors, the personalized needs of users in the future time period are predicted, and the demand prediction results are determined.

[0036] User interaction data and scenario factor data are acquired from multi-source data platforms. Data cleaning techniques are used to remove noisy data, resulting in a structured dataset. Using this structured dataset, a pre-established user profile model is employed to extract user emotional features and determine the distribution of user emotions. If the fluctuation range of user emotion distribution within a specified time period exceeds a preset threshold, the time period is dynamically divided to obtain time segmentation results. Based on the time segmentation results and scenario factor data, a random forest algorithm is used to analyze the changing trends of emotion distribution, obtaining emotion change trends. Using emotion change trends and scenario factor data, a personalized user demand prediction model is constructed, and a logistic regression algorithm is used to predict user needs, obtaining demand prediction results. If the confidence level of the demand prediction results is lower than a preset threshold, a Bayesian network is used to adjust the prediction model, combining historical data and scenario factors to obtain optimized prediction results. Based on the optimized prediction results, a personalized user demand report is generated, determining the final demand prediction result.

[0037] Specifically, by dynamically analyzing the distribution of emotional tendencies over a specific time period using a user sentiment profiling model, we can first collect users' social media data and app usage records from the past 30 days. Assuming that the daily generated sentiment data includes three categories—positive, negative, and neutral—with proportions of 40%, 30%, and 30%, respectively, we can then use sentiment analysis algorithms such as Support Vector Machines (SVM) to score the sentiment polarity of the text content. Combined with timestamp data, we can categorize sentiment tendencies by hourly time slots. For example, during work hours (8:00-17:00), the proportion of negative sentiment rises to 35%, while in the evening (18:00-22:00), the proportion of positive sentiment increases to 45%. Next, considering time period segmentation and contextual influencing factors, such as the difference between weekdays and weekends, the analysis reveals that negative sentiment peaks at 38% on Mondays, while positive sentiment increases to 50% on weekends, presumably related to work stress and leisure activities. Further time series analysis algorithms, such as the ARIMA model, can predict user sentiment tendencies over the next 7 days, concluding that negative sentiment during Monday work hours is likely to remain at 36%, while positive sentiment in the evening may drop to 42%. Based on this, personalized user needs are predicted. For example, pushing stress-relieving content on weekdays is expected to increase meditation app downloads by 20%, while pushing entertainment activities on weekends is expected to increase participation by 15%. Finally, the demand prediction results are validated using machine learning models (such as random forests), and the error rate is controlled within 5% by combining historical data. This determines that users are more inclined to use stress-relieving and entertainment services in the future, and the system automatically generates a personalized recommendation list, ensuring that the click-through rate of recommended content increases by more than 10%. Through the above analysis, a complete logical chain is formed from data collection to sentiment prediction to demand recommendation, relying on algorithms and numerical analysis to ensure the accuracy of the results.

[0038] Step S106: Based on the demand forecast results, generate a service configuration adjustment plan that matches the user's emotional tendencies, fine-tune the service content according to the real-time changes captured by dynamic data, and obtain the adjusted service parameters.

[0039] The system receives real-time captured user sentiment data and compares it against a pre-established sentiment classification rule base. Features are extracted for at least one sentiment category to obtain sentiment intensity values ​​and sentiment labels, resulting in a preliminary sentiment judgment. Based on the sentiment labels in the sentiment judgment result and combined with service usage records in dynamic data, a preset threshold range is used for matching. If the sentiment intensity value exceeds the threshold range, configuration adjustment logic is triggered to determine the direction of service parameter adjustment. According to the service parameter adjustment direction, a configuration scheme matching the sentiment is retrieved from a preset service content template library. Fine-tuning is performed based on real-time changes in data, and it is determined whether the adjusted parameters meet a preset balance standard, resulting in the final parameter configuration combination. Using this parameter configuration combination, a data mapping tool applies the adjusted parameters to the service content distribution process, obtaining real-time feedback data to determine whether further fine-tuning of parameter values ​​is needed, ultimately determining the final presentation format of the service content.

[0040] Specifically, during the implementation of service configuration adjustments, user data was first analyzed using a demand forecasting model. Assuming the system collected user service request data from the past 30 days, it was found that the peak request volume was 5000 times per day, and it was predicted that the peak volume might increase to 5500 times in the next 7 days. An ARIMA model based on time series was used for prediction, and the calculated prediction error range was ±5%, thus determining the reliability of the prediction value to be 95%. Next, combined with user sentiment analysis, natural language processing technology was used to assign sentiment scores to user feedback text. Assuming the score range is -1 to 1, the current average user sentiment score is 0.3, indicating a predominantly positive sentiment. However, some negative feedback focused on excessively long response times. The system automatically extracted the keywords "waiting" and "slow," accounting for 20%. Through correlation analysis between sentiment and keywords, it was determined that response time needs optimization. Subsequently, based on real-time changes captured from dynamic data, the system monitors service response time in real time. It finds the current average response time to be 3.2 seconds, higher than the target value of 2.5 seconds. Combined with the predicted increase in request volume, the system automatically triggers a fine-tuning mechanism, invoking a load balancing algorithm to increase the server resource allocation ratio from 60% to 75%, while simultaneously adjusting the cache hit rate target from 80% to 85%. Simulation tests show that the adjusted response time is expected to decrease to 2.4 seconds. Finally, the system obtains the adjusted service parameters, generates a new configuration scheme, including increasing the number of server cluster nodes from 10 to 12 and the bandwidth limit from 100Mbps to 120Mbps. This is automatically pushed to the server via API, completing the configuration update. The entire process forms a closed-loop logic, ensuring seamless integration of prediction, sentiment analysis, and real-time adjustments, thus improving user experience.

[0041] Step S107: The service process of the guest room management system is updated in real time through the adjusted service parameters. Combined with the feedback data of service configuration adjustment, the user emotion profile model is continuously optimized to obtain the optimized emotion cycle change pattern.

[0042] By retrieving service parameter adjustment records from the guest room management system, the impact on the service process is analyzed to obtain preliminary service response change trends. Based on these trends, and combined with the configuration feedback data stream, a preset threshold is used for comparison to determine if abnormal fluctuations exist and identify service links requiring optimization. If abnormal fluctuations are detected, relevant emotional expression features are extracted from user emotional feedback data to obtain the distribution pattern of emotional fluctuations. Analysis of the emotional fluctuation distribution pattern is input into a pre-established support vector machine model to deduce the potential change patterns of user emotional cycles. Based on the derived emotional cycle change patterns, the parameter configuration of the user emotional profile model is adjusted, resulting in an updated user emotional profile. By comparing the updated emotional profile results with historical data, the consistency and deviation are analyzed to determine the direction and extent of model optimization. Based on the direction and extent of model optimization, the service parameter configuration strategy is dynamically adjusted to generate a service process improvement plan for the guest room management system.

[0043] Specifically, by adjusting the service parameters, the service process of the guest room management system is updated in real time. The system first automatically adjusts the service response time and resource allocation based on historical data and real-time user feedback.

[0044] For example, when the system detected a 30% increase in room service requests during peak hours (18:00-21:00 daily), it adjusted the waiter allocation ratio from 1:10 to 1:8 using a dynamic scheduling algorithm (such as weighted round-robin). This reduced the response time from 5 minutes to 3.5 minutes. The algorithm calculates the optimal path based on waiter location and task saturation (saturation threshold set to 0.8), reducing unnecessary movement distance by 15%. Combining feedback data from service configuration adjustments, the system collects user ratings (1-5 points) and text comments, using natural language processing (BERT model) to analyze sentiment. Scores above 4.2 indicate positive sentiment, while scores below 3.0 indicate negative sentiment, achieving an accuracy rate of 92%. With approximately 500 feedback data points daily, the system uses an incremental update mechanism to fine-tune the sentiment analysis model hourly, improving the F1 score from 0.85 to 0.89. The system continuously optimizes the user emotion profile model. Using time series analysis (ARIMA model, p=2, d=1, q=1), combined with user stay duration (average 3.2 days) and consumption behavior (average daily spending of 500 yuan), it predicts emotion cycle changes. The system found that the probability of negative emotions increases by 20% after 18:00 on the second day of stay. Based on this, service strategies are adjusted, such as increasing the frequency of proactive greetings during this period (from once every 4 hours to once every 2 hours), improving user satisfaction by 12%. Through these steps, the system forms a closed-loop optimization logic, with service process adjustments and emotion model updates mutually reinforcing each other. The emotion cycle prediction error is controlled within ±5%, achieving precise service.

[0045] Step S108: If the optimized emotional cycle change pattern deviates from the actual user feedback, then by collecting long-term data records again, updating the voiceprint feature matrix and emotional tendency distribution, new periodic fluctuation characteristics are determined.

[0046] The system acquires long-term voice recordings from users through a data acquisition module to construct an initial voiceprint feature dataset. Key voiceprint parameters are extracted from this dataset to build a voiceprint feature matrix, which is then stored in a pre-defined database. A support vector machine (SVM) algorithm is used to classify emotional tendencies based on the voiceprint feature matrix, yielding a preliminary emotional tendency distribution. If the emotional tendency distribution deviates from the user's feedback records, a long-term data acquisition mechanism is triggered to update the voiceprint feature matrix and recalculate the emotional tendency distribution. Based on the updated emotional tendency distribution, the system analyzes periodic fluctuation characteristics to determine the changing trends of these fluctuations. By analyzing these trends and combining them with historical data records, a new emotional cycle pattern is derived and stored in the system database. After acquiring the new emotional cycle pattern, parameters are adjusted using a pattern optimization module to determine the optimized emotional cycle characteristics.

[0047] Specifically, after optimizing the emotional cycle change pattern, if deviations are found from actual user feedback, adjustments and updates can be made through information technology. First, the system automatically triggers a long-term data collection mechanism, using a smart voice device to record user voice data continuously for 30 days, 24 hours a day, with user authorization, generating an audio dataset with a total duration exceeding 720 hours. Background noise is removed through preprocessing algorithms to ensure a signal-to-noise ratio of over 20dB. Next, the system uses deep learning models such as convolutional neural networks (CNNs) to extract voiceprint features, segmenting the audio data into 5-second segments, calculating the Mel-frequency cepstral coefficients (MFCCs) of each segment, forming a 13×100 feature matrix, and performing principal component analysis (PCA) to reduce the dimensionality of the matrix, retaining 95% of the variance information to generate a new voiceprint feature matrix. Subsequently, based on the updated feature matrix, the system analyzes the distribution of emotional tendencies using a Hidden Markov Model (HMM), calculating the probability of the user's daily emotional state over the past 30 days (e.g., 60% positive, 30% neutral, and 10% negative). It then uses time series analysis (such as an ARIMA model) to predict emotional fluctuation trends, finding a cyclical fluctuation characteristic of an emotional peak every 7 days, with a peak emotional intensity of 0.85 (out of 1.0). Finally, the system compares the newly calculated cyclical fluctuation characteristics with historical data. If the deviation exceeds 10%, it automatically updates the emotional cycle model in the database and adjusts the emotional tone of the pushed content through related business modules, such as the intelligent recommendation system. For example, it increases the proportion of positive emotional content to 70% to match the user's current emotional cycle and ensure recommendation accuracy. Through this fully automated process, the system can dynamically adapt to changes in user emotions, forming a data-driven closed-loop optimization logic.

[0048] Step S109: Based on the new periodic fluctuation characteristics, repeatedly execute the process of building the emotion profile model and predicting personalized needs, continuously improve the targeting of service configuration adjustments, and obtain the final optimized service quality improvement plan.

[0049] A periodic fluctuation feature extraction algorithm is used to obtain periodic fluctuation features from user interaction data, determining the fluctuation period and amplitude pattern. Feature analysis techniques are employed to classify and cluster the obtained periodic fluctuation features, resulting in an initial feature set for the emotion profile. If the classification accuracy of the initial feature set is lower than a preset threshold, the features are re-selected using a random forest algorithm to obtain an optimized emotion profile feature set. Based on the optimized emotion profile feature set, an emotion profile model is constructed and trained using a support vector machine algorithm to obtain an emotion profile classification model. Using the emotion profile classification model and combined with historical user behavior data, personalized needs are predicted, yielding personalized need prediction results. Based on these personalized need prediction results, service configuration parameters are adjusted to generate a targeted optimized service configuration plan. If the deviation between the feedback data of the service configuration plan and the expected quality improvement target exceeds a preset threshold, the emotion profile model and the need prediction process are iteratively updated to obtain the optimal service quality improvement plan.

[0050] Specifically, based on the novel periodic fluctuation characteristics, the model first extracts the periodic features of the sentiment profile through time series analysis. The Fast Fourier Transform (FFT) algorithm is used to process user interaction data, assuming the dataset contains 10,000 user sentiment scores (range 0-10) over a 30-day period. FFT analysis reveals that the main periods are 7 days and 14 days, reflecting the weekly and bi-weekly fluctuations of user sentiment. Next, a sentiment profile model is constructed, utilizing a Long Short-Term Memory (LSTM) network to process the time series data. The input is a standardized sentiment score sequence, with 128 hidden layers, training for 100 epochs, using the Adam optimizer with a learning rate of 0.001 and a mean squared error (MSE) loss function. The model outputs a dynamic profile of user sentiment with a prediction accuracy of 85%. Subsequently, personalized demand prediction is performed, based on a collaborative filtering algorithm combined with user historical behavior.

[0051] For example, based on 1000 service interaction records, user preference vectors are calculated, and cosine similarity is used to assess user similarity with a threshold of 0.8 to predict potential user demand for the service (e.g., 60% of users prefer a response time of <2 seconds). To continuously improve service configuration, a reinforcement learning algorithm (Q-learning) is used, with the service quality score (0-100) as the reward function. Service parameters are adjusted (e.g., optimizing server response time from 2.5 seconds to 1.8 seconds). After 1000 iterations, the Q-value converges, and the service quality score improves by 15%. The final optimization scheme is validated through A / B testing. 1000 users are divided into two groups: the optimization group has a response time of 1.8 seconds, and the control group has a response time of 2.5 seconds. The results show that user satisfaction in the optimization group increases by 20%, and the service interruption rate decreases to 0.5%. Through periodic analysis, model building, demand forecasting, and configuration adjustments, a closed-loop optimization logic is formed to continuously improve service quality.

[0052] The above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the invention. Those skilled in the art will understand that implementing all or part of the above-described embodiments and making equivalent changes in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A method for emotion perception and room management based on voiceprint recognition, characterized in that, The method comprises: Real-time recording of sound data of a user in a guest room management scene by a voiceprint collection device to obtain an original voiceprint data set; preliminary classification of sound signals under different time periods and environmental conditions to obtain a preliminarily classified voiceprint signal set for subsequent emotional state analysis and feature extraction; For the preliminarily classified voiceprint signal set, a voiceprint feature extraction technology is used to deeply process each sound signal to separate out emotion-related acoustic feature parameters, construct a multi-dimensional voiceprint feature matrix containing pitch, volume and speech rate, and determine emotional feature basic data; According to the voiceprint feature matrix, a time series analysis method is used to mine the regularity of the long-term recorded emotional feature basic data to identify the emotional periodical change rule of the user in different time periods and determine the time nodes and change trends of the periodical fluctuations, wherein the time series analysis method comprises an autoregressive moving average model or a Fourier transform method; If the identified periodical fluctuation time nodes and change trends have statistically significant rules, a user emotional portrait model is constructed for these rules, the long-term data record is combined with scene influencing factors to obtain the emotional tendency distribution of the user in a specific time period; Through the user emotional portrait model, the emotional tendency distribution in a specific time period is dynamically analyzed, and the time period division and scene influencing factors are combined to predict the personalized demand of the user in a future time period and determine a demand prediction result; According to the demand prediction result, a service configuration adjustment scheme matched with the emotional tendency of the user is generated, the service content is fine-tuned according to the real-time changes captured by the dynamic data, and adjusted service parameters are obtained; Through the adjusted service parameters, the service process of the guest room management system is updated in real time, the feedback data of the service configuration adjustment are combined, the user emotional portrait model is continuously optimized, and an optimized emotional periodical change rule is obtained; If the optimized emotional periodical change rule deviates from the actual user feedback, the voiceprint feature matrix and the emotional tendency distribution are updated by re-collecting long-term data records to determine new periodical fluctuation characteristics; According to the new periodical fluctuation characteristics, the construction of the emotional portrait model and the personalized demand prediction process are cyclically executed to continuously improve the pertinence of the service configuration adjustment and obtain a final optimized service quality improvement scheme.

2. The method of claim 1, wherein the voiceprint-based emotion perception and guest room management method is characterized by, The real-time recording of sound data of a user in a guest room management scene by a voiceprint collection device, the preliminary classification of sound signals under different time periods and environmental conditions, the obtaining of an original voiceprint data set for subsequent emotional state analysis and feature extraction, and the obtaining of a preliminarily classified voiceprint signal set comprise: Recording the sound signals of the user in the guest room management scene by the voiceprint collection device and obtaining a first voiceprint data set by using a microphone array technology; According to the first voiceprint data set, processing the environmental noise by using a filter to obtain a second voiceprint data set; If the signal-to-noise ratio of the second voiceprint data set is higher than a preset threshold, classifying the second voiceprint data set according to the time period and the environmental condition by using a classification algorithm to obtain a first classified signal set; Extracting frequency and amplitude features from the first classification signal set, using fast Fourier transform technology to process the first classification signal set, obtaining a first feature set and storing it in the database; For the first feature set, match with the preset emotion state template to determine the emotion category corresponding to the first feature set; According to the emotion category and the first feature set, generate a classification report and store it in the database to determine the final voiceprint signal analysis result. 3.The emotion perception and guest room management method based on voiceprint recognition according to claim 1, wherein, The voiceprint feature extraction technology is used to deeply process each voice signal after preliminary classification, and the emotion-related acoustic feature parameters are separated from it to construct a voiceprint feature matrix containing tone, volume and speech speed multi-dimensional features, and determine the emotion feature basic data, including: Obtain initial data from voiceprint signals, separate emotion-related acoustic features using feature extraction tools to obtain a preliminary acoustic feature set; According to the preliminary acoustic feature set, quantitatively process the tone parameters, volume changes and speech rhythm to determine the corresponding multi-dimensional feature data; Through the multi-dimensional feature data, a feature matrix containing tone parameters, volume changes and speech rhythm is constructed to obtain a comprehensive feature matrix structure; If the feature value of a certain dimension in the comprehensive feature matrix structure deviates from the preset threshold range, the dimension data is standardized to obtain an adjusted feature matrix; According to the adjusted feature matrix, use support vector machine algorithm to classify and process emotion data to determine the feature distribution related to emotion; For the classified and processed emotion data, obtain the mapping relationship corresponding to the basic data to determine the final emotion feature basic data set; Through the final emotion feature basic data set, verify it combined with the pre-established acoustic feature model to obtain the verified emotion feature data.

4. The method of claim 1, wherein the voiceprint-based emotion perception and guest room management method is characterized by, According to the voiceprint feature matrix, use time series analysis method to regularly mine long-term recorded emotion feature basic data, identify the emotion periodical change rule of the user in different time periods, determine the time node and change trend of periodic fluctuations, including: Obtain voiceprint feature data to construct an initial feature matrix, which contains long-term recorded voiceprint feature data; Use standardization method to normalize the initial feature matrix to obtain a processed feature matrix; According to the processed feature matrix, apply time series analysis method to decompose the feature distribution of different time periods to obtain the fluctuation sequence of emotion features; Through the fluctuation sequence, combined with the preset threshold range, determine the periodic change rule of emotion features, if the fluctuation sequence exceeds the threshold range in a certain time period, mark the time period as a significant fluctuation interval to obtain a significant fluctuation time node set; According to the significant fluctuation time node set, use the sliding window method to analyze the time interval and fluctuation amplitude between adjacent nodes, extract the local change trend to obtain the emotion change direction in each time period; Through the local change trend, combined with the historical distribution of long-term recorded data, construct an emotion periodical fluctuation model to determine the main time node of periodic fluctuations; According to the main time node, the K-means clustering method is used to classify the emotional feature data, and the category distribution of emotional changes in different time periods is obtained; Through the category distribution, combined with the continuity characteristics of the fluctuation sequence, the conversion law between categories is analyzed, and the trend of the user's emotional periodical change is obtained.

5. The method of claim 1, wherein the voiceprint-based emotion perception and guest room management method is characterized by, If the identified periodic fluctuation time node and change trend exist statistical significant rules, then a user emotion portrait model is constructed for these rules, the long-term data record is combined with the scene influencing factors, and the emotional tendency distribution of the user in a specific time period is obtained, including: Extracting the related data of periodic fluctuations and time nodes from historical records to construct an initial data set; Using preprocessing means to remove noise data in the initial data set to obtain a cleaned fluctuation data set; According to the cleaned fluctuation data set, the relationship between the time node and the specific time is analyzed to determine whether there is a significant periodic rule. If the fluctuation data presents a repeated pattern in a specific time period, it is determined as periodic rule data; For the periodic rule data, a feature matrix is constructed to fuse the relevance of influencing factors and change trends to obtain a feature-based emotional related data set; Through the feature-based emotional related data set, a support vector machine algorithm is applied to train the user model to derive the classification result of emotional tendency and obtain the preliminary emotional tendency distribution; According to the preliminary emotional tendency distribution, combined with the specific time and the data of the periodic fluctuation, the distribution result is corrected in the time dimension. If the deviation of the corrected distribution exceeds the preset threshold, the feature matrix is adjusted to obtain the corrected emotional distribution data; Through the corrected emotional distribution data, the complete emotional portrait is constructed by mapping to the user model to determine whether the emotional portrait covers all specific time periods to obtain the final user emotional portrait result; According to the final user emotional portrait result, the distribution derivation and change trend are associated to generate emotional tendency prediction data in a specific time period to complete the dynamic monitoring of the user's emotion.

6. The method of claim 1, wherein the voiceprint-based emotion perception and guest room management method is characterized by, Through the user emotion portrait model, the emotional tendency distribution in a specific time period is dynamically analyzed, combined with the time period division and the scene influencing factors, the individualized needs of the user in the future time period are predicted, and the demand prediction result is determined, including: Obtaining user interaction data and scene factor data from a multi-source data platform, using data cleaning technology to remove noise data to obtain a structured data set; Through the structured data set, the pre-established portrait model is used to extract user emotional features to determine the user emotional distribution; If the user emotional distribution changes more than the preset threshold in a specified time period, the time period is dynamically divided to obtain the time division result; According to the time division result and the scene factor data, the random forest algorithm is used to analyze the change trend of the emotional distribution to obtain the emotional change trend; Through the emotional change trend and the scene factor data, a user individualized demand prediction model is constructed, and a logistic regression algorithm is used to predict the user demand to obtain the demand prediction result; If the confidence of the demand prediction result is lower than a preset threshold, a Bayesian network is used to adjust the prediction model in combination with historical data and scene factors to obtain an optimized prediction result. Through the optimized prediction result, a user personalized demand report is generated to determine a final demand prediction result.

7. The method of claim 1, wherein the voiceprint-based emotion perception and guest room management method is characterized by, According to the demand prediction result, a service configuration adjustment scheme matched with the user emotional tendency is generated, the service content is fine-tuned for real-time changes captured by dynamic data, and adjusted service parameters are obtained, including: Receiving real-time captured user emotional tendency data, comparing according to a pre-established emotion classification rule library, extracting features for at least one emotion category, obtaining an emotional intensity value and a tendency label, and obtaining a preliminary emotional judgment result; According to the tendency label in the emotional judgment result, in combination with service usage records in dynamic data, a preset threshold range is matched, and if the emotional intensity value exceeds the threshold range, a configuration adjustment logic is triggered to determine a service parameter adjustment direction; According to the service parameter adjustment direction, a configuration scheme matched with the emotional tendency is obtained from a preset service content template library, fine-tuned for real-time change data, and whether the adjusted parameters meet a preset balance standard is judged to obtain a final parameter configuration combination; Through the parameter configuration combination, a data mapping tool is used to apply the adjusted parameters to the service content distribution process to obtain real-time feedback data, judge whether the parameter value needs to be further fine-tuned, and determine the final presentation form of the service content.

8. The method of claim 1, wherein the voiceprint-based emotion perception and guest room management method is characterized by, Through the adjusted service parameters, the service process of the room management system is updated in real time, the feedback data of the service configuration adjustment is combined to continuously optimize the user emotional portrait model, and an optimized emotional cycle change rule is obtained, including: By obtaining the adjustment record of the service parameters from the room management system, the influence degree on the service process is analyzed to obtain a preliminary service response change trend; According to the service response change trend, in combination with the data flow of the configuration feedback, a preset threshold is compared to judge whether there is abnormal fluctuation, and the service link that needs to be optimized is determined; If abnormal fluctuation of the service link is detected, relevant emotional expression features are extracted from the user emotional feedback data to obtain the distribution mode of emotional fluctuation; Through analysis of the emotional fluctuation distribution mode, the support vector machine model is input into the pre-established support vector machine model to deduce the potential change rule of the user emotional cycle; According to the deduced emotional cycle change rule, the parameter configuration of the user emotional portrait model is adjusted to obtain an updated user emotional portrait result; By comparing the emotional portrait update result with the historical data, the consistency and deviation degree are analyzed to judge the direction and amplitude of model optimization; According to the direction and amplitude of model optimization, the configuration strategy of the service parameters is dynamically adjusted to generate a service process improvement scheme for the room management system.

9. The method of claim 1, wherein the voiceprint-based emotion perception and guest room management method is characterized by, If the optimized emotional cycle change rule deviates from the actual user feedback, long-term data records are collected again to update the voiceprint feature matrix and emotional tendency distribution, new periodic fluctuation features are judged, including: An initial voiceprint feature dataset is constructed by acquiring long-term voice records of a user through a data acquisition module; Key voiceprint parameters are extracted from the voiceprint feature dataset to construct a voiceprint feature matrix, which is stored in a preset database; The voiceprint feature matrix is classified by a support vector machine algorithm to obtain a preliminary emotion tendency distribution result; If the emotion tendency distribution result deviates from the user feedback record, a long-term data acquisition mechanism is triggered to update the voiceprint feature matrix and recalculate the emotion tendency distribution; According to the updated emotion tendency distribution, the periodic fluctuation characteristics are analyzed to determine the change trend of the fluctuation characteristics; According to the updated emotion tendency distribution, the periodic fluctuation characteristics are analyzed to determine the change trend of the fluctuation characteristics; After obtaining the new emotion cycle rule, the parameter adjustment is performed on the rule optimization module to determine the optimized emotion cycle characteristics.

10. The method of claim 1, wherein the voiceprint-based emotion perception and guest room management method is characterized by, According to the new periodic fluctuation characteristics, the construction of the emotion portrait model and the personalized demand prediction process are cyclically executed to continuously improve the pertinence of service configuration adjustment and obtain the final optimized service quality improvement scheme, including: Periodic fluctuation characteristics are obtained from user interaction data through a periodic fluctuation characteristic extraction algorithm to determine the fluctuation period and amplitude mode; Feature analysis techniques are used to classify and cluster the obtained periodic fluctuation characteristics to obtain an initial feature set of the emotion portrait; If the classification accuracy of the initial feature set is lower than a preset threshold, the features are reselected through a random forest algorithm to obtain an optimized emotion portrait feature set; According to the optimized emotion portrait feature set, an emotion portrait model is constructed, and a support vector machine algorithm is used for training to obtain an emotion portrait classification model; Through the emotion portrait classification model, combined with user historical behavior data, personalized demands are predicted to obtain a personalized demand prediction result; The personalized demand prediction result is used to adjust the service configuration parameters to generate a pertinence-optimized service configuration scheme; If the feedback data of the service configuration scheme deviates from the expected quality improvement target by more than a preset threshold, the emotion portrait model and the demand prediction process are iteratively updated to obtain an optimal service quality improvement scheme.

Images