Pedestrian street tourist emotion experience evaluation method based on wearable device and machine learning

Abstract

The invention discloses a pedestrian street tourist emotion experience evaluation method based on wearable equipment and machine learning, and belongs to the technical field of intelligent tourism data analysis. Firstly, tourist real-time data is acquired to generate a data set; secondly, preprocessing the data set, separating an emotion awakening related signal, and performing baseline standardization to recognize a high point of emotion awakening; and thirdly, combining the physiological change value with the real-time position data through time synchronization to generate an emotion map. And 4, forming a street environment element library, and quantifying each environment variable. And finally, analyzing the influence of street environment elements on tourist emotion awakening by using machine learning. And outputting the importance score of each environment element, and determining the high-importance environment element. According to the invention, the emotional experience of the tourist in the pedestrian street can be accurately evaluated in real time, and the limitation that the traditional method depends on subjective evaluation is overcome; the key environmental factors influencing the mood of the tourist can be accurately recognized through the machine learning model, and the method has high generalizability.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent tourism data analysis, and relates to a method for evaluating the emotional experience of pedestrian street tourists based on wearable devices and machine learning. In particular, it relates to an evaluation method that uses wearable sensors to collect tourists' physiological signals in real time and combines machine learning models to analyze the impact of the pedestrian street environment on tourists' emotions. Background Art

[0002] Tourists' emotional experiences on pedestrian streets are a direct factor influencing their overall travel experience. Studying tourists' emotions during their strolls is crucial for understanding tourist behavior and enhancing their travel experience. Sichuan Tourism College (Chinese invention patent CN119090571A) has disclosed a dynamic optimization method for tourism service quality based on tourists' subjective evaluations. By integrating tourist evaluation statistics within tourist attractions and evaluating service categories, this method generates tourist satisfaction information. However, this emotional evaluation method primarily relies on subjective questionnaires, which compromises the objectivity of the analysis results. The Chinese Academy of Environmental Sciences (Chinese invention patent CN117152677A) has disclosed a monitoring system for assessing tourist satisfaction at scenic spots. By collecting tourist facial image data, route duration data, and environmental data on temperature, humidity, light intensity, and ambient air quality around the route, the system generates a satisfaction rating for the scenic area. This demonstrates the potential for objective data collection to directly reflect tourist behavior and environmental experience. In recent years, wearable physiological sensors have become increasingly important in emotional assessment due to their ability to accurately and real-timely record individuals' emotional experiences within their environment. The paper [Shoval N et al., "Real-time measurement of tourists' objective and subjective emotions in time and space". Journal of travel research, 2018, 57(1): 3-16] combines objective physiological emotions, spatiotemporal data with subjective semantic information, and uses wearable sensing devices to study the temporal and spatial experiences of 68 tourists in Jerusalem along a specific walking route, providing a comprehensive approach to understanding tourists' experiences.

[0003] In summary, the generation of tourist emotions is essentially a dynamic process of individual evaluation of environmental events. Combining physiological measurements with subjective assessments can better analyze tourists' emotional experiences. Furthermore, the emotions expressed by tourists toward a particular scenic spot are necessarily the result of the combined effects of the attraction itself and multiple environmental factors surrounding it. Therefore, it is necessary to develop a tourist emotional experience assessment method that integrates subjective and objective data with multiple environmental factors, quantifying the weight of each environmental factor, and providing a scientific basis for the precise allocation of tourism resources and improving the management of scenic spots. Summary of the Invention

[0004] To address the challenges of existing technologies, this paper proposes a method for assessing the emotional experience of pedestrian street visitors based on wearable devices and machine learning. This method uses wearable sensors to collect real-time physiological data and GPS location data from visitors, and combines this with a machine learning model to analyze the impact of various environmental factors on the pedestrian street's emotions.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A method for evaluating the emotional experience of pedestrians on a pedestrian street based on wearable devices and machine learning includes the following steps:

[0007] Step 1: Obtain real-time data of tourists walking on the pedestrian street and generate a dataset. The details are as follows:

[0008] Step 1.1: Select a representative pedestrian street as the research area, ensuring that the pedestrian street has diverse environmental elements so as to comprehensively evaluate the impact of different environmental elements on tourists' emotional experience;

[0009] In step 1.2, several tourists were recruited as participants. Each participant was equipped with a wearable sensor, a GPS device, and a smartphone. The wearable sensor collected physiological data, including both emotionally aroused and relaxed baseline data. The GPS device recorded the tourists' real-time location data. The smartphone was used to collect photos of the street scene. As the tourists strolled along the pedestrian street, the devices continuously recorded real-time data to generate a dataset.

[0010] Step 2: Preprocess the physiological data in the dataset obtained in step 1 to isolate the emotional arousal-related signals, perform baseline normalization, and identify the peak points of emotional arousal. The details are as follows:

[0011] Step 2.1, preprocessing includes the following steps: cleaning and filtering the collected physiological data, including high-pass filtering and low-pass filtering to remove baseline drift, so as to reduce or eliminate noise and interference, make the signal smoother and clearer, and obtain a cleaned signal; then, the cleaned signal is separated to obtain the basic signal that is not related to the emotional changes caused by external stimuli, and the separation adopts the emotion-related signal extraction method, and after separation, it is removed to obtain the emotion arousal-related signal.

[0012] The emotion arousal-related signal reflects the autonomic nervous system response that is closely related to emotion, and its waveform changes can show the physiological activities during emotion arousal.

[0013] Step 2.2, baseline normalization: By subtracting the baseline data in the relaxed state from the physiological data in the emotional stimulation state obtained in step 1.2, the actual physiological change value of the subject after the emotion is induced is obtained, thereby eliminating individual static differences and highlighting the dynamic changes caused by emotional stimulation; finally, the peak value of the physiological change value is detected to identify the high point of emotional arousal.

[0014] The peak typically represents the peak of the physiological response caused by emotional arousal and can better reflect the immediate feedback of stimuli in the environment.

[0015] Step 3: Combine the physiological change values ​​preprocessed in step 2 with the real-time location data obtained from GPS in step 1.2 through time synchronization, eliminate invalid data, and finally generate an emotion map.

[0016] The invalid data includes data points marked as invalid due to equipment failure, signal loss, or subjective evaluation by tourists.

[0017] The proposed emotional map is based on the close connection between emotion and place. By combining physiological changes with real-time location data, the intensity of emotional arousal can be mapped onto a spatial map, displaying subjective, qualitative, and spatial information about the environment. This provides an effective method for studying the correlation between tourists' emotional experiences and street environments from a multidimensional perspective.

[0018] Preferably, the emotion map calculates the spatial distribution of emotion arousal hotspots by kernel density estimation method, reflecting the intensity of tourists' emotional experience in different areas of the pedestrian street. Specifically, the kernel density estimation method uses Gaussian kernel function to calculate the probability density distribution of emotion arousal hotspots. It is calculated specifically by the following formula (1):

[0019]

[0020] Where n is the number of hotspots, h is the bandwidth selected for the experiment, which requires fine-tuning through continuous adjustment and experimentation. In this study, h was set to 6.7m. K is the Gaussian kernel function, and di is the data dimension.

[0021] Step 4: Create a street environment factor database and quantify each environmental variable.

[0022] The street environment element database is divided into two categories: street space elements and tourist perception characteristics. Street environment elements that affect tourists' emotional experience are selected, including green index, street continuity, interface transparency, facility completeness, street layout and function, visual richness, pleasant cleanliness, humanized scale, historical atmosphere and sense of security and other environmental variables.

[0023] The Delphi method was used to organize multiple urban planning experts to conduct multiple rounds of scoring, evaluate and quantify each environmental variable, and form an environmental factor data set.

[0024] Step 5: Use machine learning to analyze the impact of street environmental factors on tourists' emotional arousal; output the importance score of each environmental factor and identify the key factors that have the greatest impact on tourists' emotional arousal; after the model training is completed, further determine the high-importance environmental factors. The details are as follows:

[0025] Step 5.1: Input environmental element data: The environmental variable data extracted from the street environmental element library generated in Step 4 is used as the independent variable, and the emotional arousal intensity data extracted from the emotion map generated in Step 3 is used as the target variable. Emotional arousal intensity is quantified by peak detection results of physiological signals, indicating the level of emotional arousal of tourists at different environmental nodes.

[0026] In step 5.2, normalize the environmental factor data and emotional arousal data to obtain the final dataset, eliminating dimensional differences between variables and ensuring effective training of the machine learning model. The final dataset is then divided proportionally into a training set and a test set. The training set is used for machine learning model training, and the test set is used for performance evaluation of the machine learning model.

[0027] In Step 5.3, Extreme Gradient Boosting Regression (XGBR) was selected as the machine learning model due to its advantages in handling mixed data types, nonlinear relationships, and feature importance assessment. The XGBR machine learning model was trained using the training set data. The model iteratively optimized the objective function to gradually fit the relationship between environmental factors and emotional arousal. A cross-validation approach was used to divide the training set into multiple subsets, with some subsets used for training and the remaining subsets used for validation to ensure the model's generalization ability. The regression coefficient R², root mean square error (RMSE), and feature importance score were output. After multiple iterations of training and cross-validation, the model's R² and RMSE met the expected standards, demonstrating that the model has strong explanatory power and predictive accuracy for the relationship between environmental factors and emotional arousal.

[0028] The feature importance score is calculated based on the frequency of use of the feature in the machine learning model and its contribution to the target variable obtained in step 5.2. Specifically, it refers to the weight obtained after model training and is calculated using the following formula (2):

[0029]

[0030] Among them, Importance i is the importance score of the i-th feature, T is the total number of trees in the model, is the gain of the i-th feature in the t-th tree.

[0031] Step 5.4: Sort all street environment elements according to their feature importance scores. The ranking is performed from high to low based on feature importance scores, with higher scores indicating a greater impact on tourists' emotional arousal. The mean and standard deviation of the feature importance scores for all street environment elements are calculated. Features with feature importance scores above a preset threshold are selected as high-importance environmental elements.

[0032] The preset threshold is usually the mean of the feature importance score plus one standard deviation, which is calculated by the following formula (3):

[0033]

[0034] Where μ is the mean of all feature importance scores and σ is the standard deviation.

[0035] If the characteristic importance scores of all street environmental elements are lower than the preset threshold, the street environmental elements corresponding to the first three characteristic importance scores are selected as high-importance environmental elements.

[0036] Through the threshold screening, it can be ensured that the selected street environment elements have a significant impact on tourists' emotional arousal.

[0037] Finally, in step 5.5, after selecting high-importance environmental factors using the above rules, domain knowledge is further incorporated to eliminate features with unclear practical significance or irrelevant to the target variable, ultimately determining a set of high-importance features. This domain knowledge includes theoretical and practical experience in related fields such as pedestrian street environmental design and tourist behavior research. Through these steps, the method can scientifically and systematically identify high-importance features, providing guidance for pedestrian street environmental optimization.

[0038] The beneficial effects of the present invention are:

[0039] (1) By combining wearable sensors and machine learning technology, this invention can accurately and in real time evaluate tourists’ emotional experience in pedestrian streets, overcoming the limitations of traditional methods that rely on subjective evaluation;

[0040] (2) This invention uses machine learning models to provide data support, which can accurately identify key environmental factors that affect tourists' emotions and provide targeted optimization suggestions for urban planners and decision makers;

[0041] (3) The method of the present invention has high scalability and is applicable to the emotional experience evaluation of other similar tourist attractions and public spaces. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A framework diagram of the method of the present invention;

[0043] Figure 2This is a diagram of the data preprocessing process; Figure 2 (a) is the physiological signal EDA-Raw; Figure 2 (b) is the signal after cleaning EDA-Clean; Figure 2 (c) in the figure is the basic signal EDA-Tonic; Figure 2 (d) is the emotional arousal related signal EDA-Phasic;

[0044] Figure 3 It is a peak diagram of an example of the present invention;

[0045] Figure 4 This is the emotion-space mapping process in the example of the present invention; Figure 4 (a) shows the node distribution in a high-wake-up environment in an embodiment of the present invention; Figure 4 (b) is the emotion map in the embodiment of the present invention;

[0046] Figure 5 A flowchart for machine learning;

[0047] Figure 6 is the machine learning result in the example of the present invention; Figure 6 (a) is the XGBR prediction result in an embodiment of the present invention; Figure 6 (b) in the figure is the result of ranking the importance of environmental factor features in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] To facilitate understanding of the present invention, the present invention is described in more detail below with reference to the accompanying drawings and specific embodiments. The accompanying drawings provide preferred embodiments of the present invention. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in this specification. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0049] To address the challenges of existing technologies, this paper proposes a method for assessing the emotional experience of pedestrian street visitors based on wearable devices and machine learning. This method uses wearable sensors to collect real-time physiological data and GPS location data from visitors, and combines this with a machine learning model to analyze the impact of various environmental factors on the pedestrian street's emotions.

[0050] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0051] A method for evaluating the emotional experience of pedestrian street tourists based on wearable devices and machine learning. Figure 1 The framework diagram of the method of the present invention includes the following steps:

[0052] Step 1: Obtain real-time data of tourists walking on the pedestrian street and generate a dataset. The details are as follows:

[0053] Step 1.1: Select a representative pedestrian street as the research area, ensuring that the pedestrian street has diverse environmental characteristics so as to comprehensively evaluate the impact of different environmental factors on tourists' emotional experience;

[0054] In step 1.2, several tourists were recruited as participants. Participants were equipped with wearable sensors, GPS devices, and smartphones. The wearable sensors collected physiological data, including both emotionally aroused and relaxed baseline data. The GPS devices recorded real-time location data. The smartphones collected photos of the street scenes. As tourists strolled along the pedestrian street, the devices continuously recorded real-time data to generate a dataset.

[0055] In the embodiment of the present invention, 45 participants were recruited to ensure the validity and representativeness of the experiment, and the sample included 22 males and 23 females.

[0056] In this embodiment of the present invention, the wearable device uses the electrodermal activity (EDA) sensor from the Biosignalsplux multi-channel physiological recorder to assess skin resistance by placing two electrode patches on the participant's middle and index fingers. The device samples data at a 1000Hz frequency and has been clinically tested to demonstrate high efficacy and accuracy.

[0057] In the embodiment of the present invention, the GPS device uses a Binghe 980 RTK GPS device, which can record geographic coordinates and timestamps at a frequency of 1 Hz.

[0058] Step 2: Preprocess the physiological data in the dataset obtained in step 1 to separate the emotional arousal related signals, perform baseline normalization, and identify the peak points of emotional arousal. Figure 2 The data preprocessing process diagram is as follows:

[0059] Step 2.1, preprocessing, includes the following steps: Cleaning and filtering the collected physiological signal EDA-Raw, including high-pass and low-pass filtering, to reduce or eliminate noise and interference, making the signal smoother and clearer, resulting in a cleaned signal EDA-Clean. Subsequently, isolating the fundamental signal EDA-Tonic, which is unrelated to emotional changes caused by external stimuli, to obtain the emotional arousal-related signal EDA-Phasic. The EDA-Phasic signal reflects the autonomic nervous system response closely related to emotion, and its waveform changes can indicate skin conductance activity during emotional arousal.

[0060] Step 2.2, baseline normalization is as follows: by subtracting the baseline data in the relaxed state from the physiological data in the emotional stimulation state obtained in step 1.2, the actual physiological change value after the subject is induced by emotion is obtained, thereby eliminating individual static differences and highlighting the dynamic changes caused by emotional stimulation; finally, the peak value of the physiological change value is detected to identify the peak point of emotional arousal. Figure 3 This is a peak diagram of an example of the present invention.

[0061] The peak typically represents the peak of the physiological response caused by emotional arousal and can better reflect the immediate feedback of stimuli in the environment.

[0062] Step 3: Combine the physiological change values ​​preprocessed in step 2 with the real-time location data obtained from GPS in step 1.2 through time synchronization, eliminate invalid data, and finally generate an emotion map. Figure 4 This is a diagram of the emotion-space mapping process in the example of the present invention. The specific operations are as follows:

[0063] Step 3.1: Combine the physiological change value pre-processed in step 2 with the real-time location data obtained from GPS in step 1.2 through time synchronization to obtain synchronized physiological data and real-time location data, ensuring that each physiological data change value corresponds to the corresponding geographic location information. Figure 4 (a) shows the node distribution in a high-awake environment according to an embodiment of the present invention.

[0064] Step 3.2, invalid data elimination: Combine the synchronized physiological data and real-time location data with the information on the photos obtained in step 1.2 to eliminate invalid data. Invalid data includes data points marked as invalid due to equipment failure, signal loss, or subjective evaluation by tourists.

[0065] Step 3.3, emotion map generation: After eliminating invalid data, the synchronized physiological data is combined with the real-time location data to generate an emotion map. Figure 4 (b) is the emotion map in an embodiment of the present invention.

[0066] Preferably, the emotion map calculates the spatial distribution of emotion arousal hotspots by kernel density estimation method, reflecting the intensity of tourists' emotional experience in different areas of the pedestrian street. Specifically, the kernel density estimation method uses Gaussian kernel function to calculate the probability density distribution of emotion arousal hotspots. It is calculated specifically by the following formula (1):

[0067]

[0068] Where n is the number of hotspots, h is the bandwidth selected for the experiment, which requires fine-tuning through continuous adjustment and experimentation. In this study, h was set to 6.7m. K is the Gaussian kernel function, and di is the data dimension.

[0069] The proposed emotional map is based on the close connection between emotion and place. By combining physiological changes with GPS data, the intensity of emotional arousal can be mapped onto a spatial map, displaying subjective, qualitative, and spatial information about the environment. This provides an effective method for studying the correlation between tourists' emotional experiences and street environments from a multidimensional perspective.

[0070] Step 4: Create a street environment factor database and quantify each environmental variable.

[0071] The street environment element database is divided into two categories: street space elements and tourist perception characteristics. Street environment elements that affect tourists' emotional experience are selected, including green index, street continuity, interface transparency, facility completeness, street layout and function, visual richness, pleasant cleanliness, humanized scale, historical atmosphere and sense of security and other environmental variables.

[0072] The Delphi method was used to organize multiple urban planning experts to conduct multiple rounds of scoring, evaluate and quantify each environmental variable, and form an environmental factor data set.

[0073] Step 5: Use machine learning to analyze the impact of street environmental factors on tourists' emotional arousal; output the importance score of each environmental factor and identify the key factors that have a great impact on tourists' emotional arousal; after the model training is completed, further determine the high-importance environmental factors. Figure 5 This is a machine learning flow chart. The details are as follows:

[0074] Step 5.1: Input environmental element data: The environmental variable data extracted from the street environmental element library generated in Step 4 is used as the independent variable, and the emotional arousal intensity data extracted from the emotion map generated in Step 3 is used as the target variable. Emotional arousal intensity is quantified by peak detection results of physiological signals, indicating the level of emotional arousal of tourists at different environmental nodes.

[0075] In step 5.2, normalize the environmental factor data and emotional arousal data to obtain the final dataset, eliminating dimensional differences between variables and ensuring effective training of the machine learning model. The final dataset is then divided proportionally into a training set and a test set. The training set is used for machine learning model training, and the test set is used for performance evaluation of the machine learning model.

[0076] In Step 5.3, Extreme Gradient Boosting Regression (XGBR) was selected as the machine learning model due to its advantages in handling mixed data types, nonlinear relationships, and feature importance assessment. The XGBR machine learning model was trained using the training set data. The model iteratively optimized the objective function to gradually fit the relationship between environmental factors and emotional arousal. A cross-validation approach was used to divide the training set into multiple subsets, with some subsets used for training and the remaining subsets used for validation to ensure the model's generalization ability. The regression coefficient R², root mean square error (RMSE), and feature importance score were output. After multiple iterations of training and cross-validation, the model's R² and RMSE met the expected standards, demonstrating that the model has strong explanatory power and predictive accuracy for the relationship between environmental factors and emotional arousal.

[0077] In this example, a 5-fold cross-validation method is used for model training and validation. For example, in the first round of cross-validation, the training dataset is split into 5 smaller datasets, including 4 training data and 1 validation data. The 4 training data are used to train the ML model, and the validation data is used to evaluate the performance of the ML model. The remaining four cross-validations are similar to the first one, resulting in 5 ML models. These five ML models are then merged into one ML model, and its performance is evaluated using R² and RMSE scores. Figure 6 (a) is the XGBR prediction result in an embodiment of the present invention.

[0078] The feature importance score is calculated based on the frequency of use of the feature in the machine learning model and its contribution to the target variable obtained in step 5.2. Specifically, it refers to the weight obtained after model training and is calculated using the following formula (2):

[0079]

[0080] Among them, Importance i is the importance score of the i-th feature, T is the total number of trees in the model, is the gain of the i-th feature in the t-th tree.

[0081] Step 5.4, sort all street environment elements according to feature importance scores. Figure 6 (b) shows the importance ranking results of environmental factors in an embodiment of the present invention. The ranking is arranged from high to low based on importance score, with higher scores indicating a greater impact on tourists' emotional arousal. The mean and standard deviation of the importance scores of all street environmental factors are calculated. Features with importance scores above a preset threshold are selected as high-importance environmental factors.

[0082] The preset threshold is usually the mean of the feature importance score plus one standard deviation, which is calculated by the following formula (3):

[0083]

[0084] Where μ is the mean of all feature importance scores and σ is the standard deviation.

[0085] If the characteristic importance scores of all street environmental elements are lower than the preset threshold, the street environmental elements corresponding to the first three characteristic importance scores are selected as high-importance environmental elements.

[0086] Through the threshold screening, it can be ensured that the selected street environment elements have a significant impact on tourists' emotional arousal.

[0087] In step 5.5, after selecting high-importance environmental factors using the above rules, domain knowledge is further incorporated to eliminate features with unclear practical significance or irrelevant to the target variable, ultimately determining a set of high-importance features. This domain knowledge includes theoretical and practical experience in related fields such as pedestrian street environmental design and tourist behavior research. Through these steps, the method can scientifically and systematically identify high-importance features, providing guidance for pedestrian street environmental optimization.

[0088] Through the above steps, the present invention provides a method for evaluating the emotional experience of pedestrian street tourists based on wearable devices and machine learning. It can collect tourists' physiological data and GPS positioning data in real time, combine machine learning models to analyze the impact of the pedestrian street environment on tourists' emotions, generate an emotional map, and identify key environmental factors, providing a scientific basis for optimizing the pedestrian street environment.

[0089] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, or two or more steps / components can be partially combined into new steps / components to achieve the purpose of the present invention.

[0090] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for evaluating the emotional experience of pedestrians on a pedestrian street based on wearable devices and machine learning, characterized by: The method for evaluating the emotional experience of pedestrian street tourists comprises the following steps: Step 1: Obtain real-time data of tourists walking on the pedestrian street and generate a dataset; specifically: Step 1.1, select the study area; In step 1.2, participants were equipped with wearable sensors, GPS devices, and smartphones. The wearable sensors were used to collect physiological data from tourists, including physiological data under emotional stimulation and baseline data under relaxation. The GPS devices were used to record the tourists' real-time location data. The smartphones were used to collect photos of the street scenes taken by tourists. As tourists strolled along the pedestrian street, the devices continuously recorded real-time data to generate a dataset. Step 2: Preprocess the physiological data in the dataset to isolate emotional arousal-related signals, perform baseline normalization, and identify the peak points of emotional arousal; the details are as follows: Step 2.1, preprocessing the physiological data in the data set obtained in step 1 to obtain emotional arousal related signals; Step 2.2, baseline normalization: Subtract the baseline data in the relaxed state from the physiological data in the emotional stimulation state obtained in step 1.2 to obtain the actual physiological changes of the subject after the emotion is induced. Finally, the peak value of the physiological change value is detected to identify the peak of emotional arousal. Step 3: The physiological change values ​​preprocessed in step 2 are combined with the real-time location data obtained from GPS in step 1.2 through time synchronization, and invalid data is eliminated to finally generate an emotion map. The emotion map calculates the spatial distribution of emotional arousal hotspots using the kernel density estimation method, reflecting the intensity of tourists' emotional experience in different areas of the pedestrian street. Step 4: Form an environmental factor dataset and quantify each street environmental factor; Step 5: Use machine learning to analyze the impact of street environmental factors on tourists' emotional arousal; output the importance score of each environmental factor and identify the key factors that have the greatest impact on tourists' emotional arousal; after model training is completed, determine the highly important environmental factors, specifically: Step 5.1, input environmental element data: The environmental variable data extracted from the street environmental element library generated in Step 4 are used as independent variables, and the emotional arousal intensity data extracted from the emotion map generated in Step 3 are used as target variables. The emotional arousal intensity is quantified by the peak detection results of physiological signals, indicating the emotional arousal level of tourists at different environmental nodes. In step 5.2, the environmental factor data and emotional arousal data are normalized to obtain the final dataset to ensure the training effect of the machine learning model. The final dataset is divided into a training set and a test set. The training set is used for machine learning model training, and the test set is used for machine learning model performance evaluation. Step 5.3: Use the training set data to train the machine learning model. The machine learning model iteratively optimizes the objective function and gradually fits the relationship between environmental factors and emotional arousal. Through multiple iterations of training and cross-validation, a trained model and corresponding feature importance scores are obtained. Step 5.4: Sort all street environmental elements according to their feature importance scores; calculate the mean and standard deviation of the feature importance scores of all street environmental elements; select features with feature importance scores above a preset threshold as high-importance environmental elements; if the feature importance scores of all street environmental elements are below the preset threshold, select the street environmental elements corresponding to the top three feature importance scores as high-importance environmental elements; In step 5.5, after filtering out high-importance environmental factors through the above rules, combine domain knowledge to determine the high-importance feature set.

2. The method for evaluating the emotional experience of pedestrians in a pedestrian street based on wearable devices and machine learning according to claim 1 is characterized in that: The preprocessing in step 2.1 is as follows: cleaning and filtering the collected physiological data, including high-pass filtering and low-pass filtering to remove baseline drift, to obtain a cleaned signal; then, separating the cleaned signal to obtain a basic signal that is not related to the emotional changes caused by external stimuli, using an emotion-related signal extraction method, and removing it after separation to obtain an emotion arousal-related signal.

3. The method for evaluating the emotional experience of pedestrians in a pedestrian street based on wearable devices and machine learning according to claim 1 is characterized in that: In step 3, invalid data includes data points marked as invalid due to equipment failure, signal loss, or subjective evaluation by tourists.

4. The method for evaluating the emotional experience of pedestrians in a pedestrian street based on wearable devices and machine learning according to claim 1 is characterized in that: In step 3, the emotion map is based on the close connection between emotion and location; by combining physiological change values ​​with real-time location data, the intensity of emotional arousal can be mapped onto a spatial map, displaying subjective, qualitative and spatial information about the environment.

5. The method for evaluating the emotional experience of pedestrians in a pedestrian street based on wearable devices and machine learning according to claim 1 is characterized in that: In step 3, the kernel density estimation method uses a Gaussian kernel function to calculate the probability density distribution of the emotional arousal hotspots; specifically, it is calculated by the following formula (1): ， Where n is the number of hotspots, h is the bandwidth selected for the experiment, which needs to be fine-tuned through continuous adjustment and experimentation; in this study, h is set to 6.7m; K is the Gaussian kernel function, and di is the data dimension.

6. The method for evaluating the emotional experience of pedestrians in a pedestrian street based on wearable devices and machine learning according to claim 1 is characterized in that: In step 4, the street environment element database is divided into two categories: street space elements and tourist perception characteristics, from which street environment elements that affect tourists' emotional experience are selected. Street environment elements include greening index, street continuity, interface transparency, facility completeness, street layout and function, visual richness, pleasant cleanliness, human scale, historical atmosphere, and sense of security; the Delphi method is used to organize multiple urban planning experts to conduct multiple rounds of scoring, and each environmental variable is evaluated and quantified to form an environmental element data set.

7. The method for evaluating the emotional experience of pedestrians in a pedestrian street based on wearable devices and machine learning according to claim 1 is characterized in that: In step 5.3, select extreme gradient boosting regression (XGBR) as the machine learning model; use the cross-validation method to divide the training set into multiple subsets, use some subsets for training in turn, and use the remaining subsets for validation; output the regression coefficient R², root mean square error (RMSE), and feature importance score.

8. The method for evaluating the emotional experience of pedestrians in a pedestrian street based on wearable devices and machine learning according to claim 7 is characterized in that: The feature importance score is calculated based on the frequency of use of the feature in the machine learning model and its contribution to the target variable obtained in step 5.

2. Specifically, it refers to the weight obtained after model training and is calculated using the following formula (2): ， Among them, Importance i is the importance score of the i-th feature, T is the total number of trees in the model, is the gain of the i-th feature in the t-th tree.

9. The method for evaluating the emotional experience of pedestrians in a pedestrian street based on wearable devices and machine learning according to claim 7 is characterized in that: In step 5.4, the higher the feature importance score, the greater the impact of the feature on tourists' emotional arousal. The preset threshold is usually the mean of the feature importance score plus one standard deviation, which is specifically calculated by the following formula (3): ， Where μ is the mean of all feature importance scores and σ is the standard deviation.

10. The method for evaluating the emotional experience of pedestrians in a pedestrian street based on wearable devices and machine learning according to claim 7, characterized in that: In step 5.5, the domain knowledge includes theoretical and practical experience in related fields such as pedestrian street environment design and tourist behavior research.

Images