Old people emotion detection method and system based on artificial intelligence
By acquiring and integrating physiological, facial images, speech and posture behavior data of the elderly, the problem of insufficient reliability and accuracy of traditional emotions detection methods is solved, and more accurate emotional state capture and detection is achieved, which is suitable for the elderly. Mental health monitoring.
Patent Information
- Application Number
- CN202411144028.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-20
- Publication Date
- 2024-11-29
AI Technical Summary
Traditional emotion detection methods rely on self-report and single physiological signals, making it difficult to accurately capture the multidimensional emotional state of the elderly, and there is insufficient reliability and accuracy.
Using an artificial intelligence-based method, physiological data, facial image data, speech data and pose behavior data of the elderly are obtained, preprocessed and feature extraction are performed, and inputted into the emotion classification module by fusing multiple data sources to generate detection results.
It significantly improves the accuracy of emotion detection, reduces the possibility of misjudgment and misjudgment, and can more comprehensively capture the emotional state of the elderly, which is suitable for mental health monitoring of the elderly.
Smart Images

Figure CN119049689A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of emotion recognition, and in particular to an artificial intelligence-based method and system for detecting emotions of the elderly. Background Art
[0002] As the global population ages, the mental health of the elderly is receiving increasing attention. The elderly often face emotional problems such as loneliness, depression, and anxiety, which not only affect their quality of life but may also lead to serious health risks. Therefore, it is particularly important to identify and monitor the emotional state of the elderly in a timely manner.
[0003] Traditional emotion detection methods mainly rely on self-reporting and simple physiological signal analysis. These methods have the following shortcomings:
[0004] Limitations of self-reporting: Older people may have difficulty expressing their emotions accurately due to cognitive impairment, decreased language ability, or sociocultural factors. This makes emotion detection methods based on questionnaires and self-assessments less reliable.
[0005] The singleness of physiological signals: Traditional methods usually rely on a single physiological signal (such as heart rate, skin galvanic response, etc.), while ignoring the diversity and complexity of emotions. Emotional states are often multidimensional, and a single signal is difficult to fully reflect the emotional changes of the elderly. Summary of the invention
[0006] In order to overcome the deficiencies of the prior art, the object of the present invention is to provide an artificial intelligence-based method and system for detecting emotions in the elderly.
[0007] To achieve the above object, the present invention provides the following solutions:
[0008] An artificial intelligence-based emotion detection method for the elderly, comprising:
[0009] Obtain physiological data, facial image data, voice data, and posture behavior data of the elderly;
[0010] Preprocessing the physiological data, the facial image data, the voice data and the posture behavior data respectively to obtain physiological preprocessing data, facial preprocessing data, voice preprocessing data and posture behavior preprocessing data; the voice preprocessing data is text data obtained by performing signal denoising and text conversion on the voice data;
[0011] Extracting features from the physiological preprocessing data, the facial preprocessing data, the voice preprocessing data, and the posture behavior preprocessing data, and fusing the extracted features to obtain fused features;
[0012] The fused features are input into the trained emotion classification module to obtain the detection results.
[0013] Preferably, the physiological data preprocessing step includes:
[0014] translating the signal of the physiological data to obtain a translated signal;
[0015] Performing wavelet decomposition on the shifted signal to obtain a plurality of wavelet coefficients;
[0016] The filtering threshold is determined according to the decomposition scale and length of the signal; the calculation formula of the filtering threshold is: Among them, λ is the filtering threshold, j is the decomposition scale of the translated signal, and d j is the decomposition of wavelet coefficients with scale j, N represents the length of the signal after translation, and median represents the median operation;
[0017] Constructing a denoising function according to the filtering threshold;
[0018] The denoising function is used to remove the noise of the translated signal to obtain the physiological preprocessing data.
[0019] Preferably, removing the noise of the translated signal by using the denoising function to obtain the physiological preprocessing data comprises:
[0020] The corresponding wavelet coefficients are removed by using the denoising function to obtain a smooth signal; wherein the denoising function is: Among them, X represents the wavelet coefficient, n represents the adjustment coefficient;
[0021] The smoothed signal is reconstructed to obtain the physiological preprocessing data.
[0022] Preferably, the signal reconstruction formula is:
[0023]
[0024] in, w j represents the energy weight of the wavelet coefficients, ||X j,k ||2 is the L2 norm of the denoised wavelet coefficients, is the reconstructed signal, J is the number of decomposition layers, ψ j',k (t) is the wavelet basis function, k is the wavelet coefficient index at a specific scale, X j,k is the kth wavelet coefficient at the jth layer.
[0025] Preferably, the preprocessing step of the facial image data comprises:
[0026] Performing filtering on the facial image data to obtain a filtered image;
[0027] Constructing an image enhancement window, and using the image enhancement window to traverse the filtered image for denoising, to obtain a denoised image;
[0028] The denoised image is subjected to histogram equalization processing to obtain the facial preprocessing data.
[0029] Preferably, the preprocessing step of the speech data includes:
[0030] Using Wiener filtering method to remove background noise of the speech data;
[0031] The data after background noise removal is converted into text to obtain the speech preprocessing data.
[0032] Preferably, the physiological preprocessing data, the facial preprocessing data, the voice preprocessing data and the posture behavior preprocessing data are subjected to feature extraction, and the extracted features are fused to obtain fused features, including:
[0033] Extracting features from the physiological preprocessing data, the facial preprocessing data, the voice preprocessing data, and the posture behavior preprocessing data to obtain physiological data features, facial image features, voice data features, and posture behavior features;
[0034] Performing feature splicing on the physiological data features, the facial image features, the voice data features and the posture behavior features to obtain a unified feature vector;
[0035] The unified feature vector is weightedly fused to obtain the fused feature.
[0036] Preferably, the physiological data features include time domain features and frequency domain features; the facial image features include geometric features and expression features; and the speech data features include MFCC features and emotion features.
[0037] Preferably, the gesture behavior features include body position and posture information, joint angle information, hand movement information, movement speed information and gesture change frequency information; the text data features include context information, grammatical feature information, vocabulary feature information, emotional information and topic information.
[0038] An artificial intelligence-based emotion detection system for the elderly, comprising:
[0039] A data acquisition module, used to acquire physiological data, facial image data, voice data, and posture behavior data of the elderly;
[0040] A preprocessing module, used to preprocess the physiological data, the facial image data, the voice data and the gesture behavior data respectively to obtain physiological preprocessing data, facial preprocessing data, voice preprocessing data and gesture behavior preprocessing data;
[0041] A fusion module, used for extracting features from the physiological preprocessing data, the facial preprocessing data, the voice preprocessing data and the posture behavior preprocessing data, and fusing the extracted features to obtain fused features;
[0042] The detection module is used to input the fusion features into the trained emotion classification module to obtain the detection results.
[0043] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:
[0044] The present invention provides an artificial intelligence-based emotion detection method and system for the elderly, the method comprising: obtaining physiological data, facial image data, voice data and posture behavior data of the elderly; preprocessing the physiological data, facial image data, voice data and posture behavior data respectively to obtain physiological preprocessing data, facial preprocessing data, voice preprocessing data and posture behavior preprocessing data; extracting features from the physiological preprocessing data, facial preprocessing data, voice preprocessing data and posture behavior preprocessing data, and fusing the extracted features to obtain fused features; inputting the fused features into a trained emotion classification module to obtain a detection result. The present invention combines physiological data, facial images, voice data and behavioral postures to comprehensively capture the emotional state of the elderly and overcome the limitations of a single data source. And by fusing multiple data sources, the accuracy of emotion detection is significantly improved, reducing the possibility of misjudgment and missed judgment. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0046] Figure 1 A flow chart of a method provided by an embodiment of the present invention;
[0047] Figure 2 A schematic diagram of the system structure provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0049] The purpose of the present invention is to provide an artificial intelligence-based emotion detection method and system for the elderly, which combines physiological data, facial images and voice data to comprehensively capture the emotional state of the elderly and overcome the limitations of a single data source.
[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0051] Figure 1 A flow chart of a method provided by an embodiment of the present invention, such as Figure 1 As shown, the present invention provides an artificial intelligence-based elderly emotion detection method, comprising:
[0052] Step 100: Acquire physiological data, facial image data, voice data, and posture behavior data of the elderly;
[0053] Step 200: Preprocessing the physiological data, the facial image data, the voice data and the posture behavior data respectively to obtain physiological preprocessing data, facial preprocessing data, voice preprocessing data and posture behavior preprocessing data; the voice preprocessing data is text data obtained by performing signal denoising and text conversion on the voice data;
[0054] Step 300: extracting features from the physiological preprocessing data, the facial preprocessing data, the voice preprocessing data, and the posture behavior preprocessing data, and fusing the extracted features to obtain fused features;
[0055] Step 400: Input the fused features into the trained emotion classification module to obtain the detection results.
[0056] Optionally, the data detection means of this embodiment is specifically as follows:
[0057] Collect physiological signals such as heart rate, blood pressure, and skin galvanic response through wearable devices (such as heart rate monitors, blood pressure monitors, etc.). Use a camera to capture the facial image of the elderly, ensuring that the shooting is done under good lighting conditions. Record through a microphone to obtain the elderly's voice samples.
[0058] Furthermore, the posture behavior data of this embodiment is obtained in the following manner:
[0059] Wearable devices (such as smart watches and fitness trackers) are used to capture motion and posture. For example, the present embodiment can also capture video data through a camera and then use computer vision technology to analyze the posture. In addition, the present embodiment also uses professional motion capture equipment to obtain accurate posture and motion data.
[0060] Furthermore, the text data corresponding to the voice data in this embodiment includes: text data obtained by performing text recognition on the voice data generated by the patient's interview record and question-and-answer record, which includes:
[0061] 1. Basic patient information: name, age, gender and other basic demographic information, contact information and emergency contact information.
[0062] 2. Lifestyle and habits: Information about the patient's lifestyle habits (such as diet, exercise, smoking, drinking, etc.) helps doctors comprehensively assess their health status.
[0063] 3. Interview records: The conversation between the doctor and the patient can be obtained through audio recording or real-time text recording, and then transcribed into text data.
[0064] 4. Medical consultation records: When patients receive online medical consultation or telemedicine services, the consultation records or chat conversations generated can be used as text data.
[0065] Through the above information, the text data of this embodiment can fully reflect the patient's health status, diagnosis results and treatment process, and help medical decision-making and the formulation of subsequent treatment plans.
[0066] Preferably, the physiological data preprocessing step includes:
[0067] translating the signal of the physiological data to obtain a translated signal;
[0068] Performing wavelet decomposition on the shifted signal to obtain a plurality of wavelet coefficients;
[0069] The filtering threshold is determined according to the decomposition scale and length of the signal; the calculation formula of the filtering threshold is: Among them, λ is the filtering threshold, j is the decomposition scale of the translated signal, and d j is the decomposition of wavelet coefficients with scale j, N represents the length of the signal after translation, and median represents the median operation;
[0070] Constructing a denoising function according to the filtering threshold;
[0071] The denoising function is used to remove the noise of the translated signal to obtain the physiological preprocessing data.
[0072] Specifically, in the wavelet threshold denoising process, the quality of threshold selection directly affects the effect of noise reduction. Since the wavelet coefficient amplitude of noise has the characteristic of decreasing with the increase of decomposition scale, the present invention introduces the decomposition scale to adaptively adjust the filtering threshold.
[0073] Preferably, removing the noise of the translated signal by using the denoising function to obtain the physiological preprocessing data comprises:
[0074] The corresponding wavelet coefficients are removed by using the denoising function to obtain a smooth signal; wherein the denoising function is: Among them, X represents the wavelet coefficient, n represents the adjustment coefficient;
[0075] The smoothed signal is reconstructed to obtain the physiological preprocessing data.
[0076] Optionally, commonly used threshold functions include hard threshold functions and soft threshold functions. Due to the discontinuity of the contraction function, the hard threshold function will produce some artificial noise points in the recovered signal, causing the reconstructed signal to oscillate; the soft threshold function shrinks the wavelet coefficients greater than the threshold. Although the processed wavelet coefficients have good continuity, some useful high-frequency information will be lost. Based on this, the present invention proposes the above-mentioned denoising function to address the shortcomings of the hard threshold function and the soft threshold function.
[0077] Furthermore, the present invention uses wavelet coefficients to measure attenuation deviation, so that the attenuation of wavelet coefficients with larger absolute values decreases as their absolute values increase, thereby effectively avoiding the loss of high-frequency information and improving the signal-to-noise ratio of the signal.
[0078] Preferably, the signal reconstruction formula is:
[0079]
[0080] in, w j represents the energy weight of the wavelet coefficients, ||X j,k ||2 is the L2 norm of the denoised wavelet coefficients, is the reconstructed signal, J is the number of decomposition layers, ψ j',k (t) is the wavelet basis function, k is the wavelet coefficient index at a specific scale, X j,k is the kth wavelet coefficient at the jth layer.
[0081] Specifically, this embodiment introduces the energy evaluation of wavelet coefficients through a weighted mechanism, thereby improving the robustness and accuracy of the reconstructed signal. The resulting smoothed signal can be better used for subsequent emotion analysis and recognition, providing effective data support for emotion detection in the elderly.
[0082] Preferably, the preprocessing step of the facial image data comprises:
[0083] Performing filtering on the facial image data to obtain a filtered image;
[0084] Constructing an image enhancement window, and using the image enhancement window to traverse the filtered image for denoising, to obtain a denoised image;
[0085] The denoised image is subjected to histogram equalization processing to obtain the facial preprocessing data.
[0086] Optionally, the original filtering algorithm of this embodiment, such as the mean filtering algorithm, performs mean processing on the pixel points in each neighborhood (i.e., the denoising window) on the image (regardless of whether there are noise points), so the image after processing will be distorted. The present invention can find the characteristic noise points on the image by utilizing the image enhancement window, and then process the corresponding characteristic noise points. It can smooth out the noise points in the image while retaining the original pixel information on the facial image as much as possible.
[0087] Preferably, the preprocessing step of the speech data includes:
[0088] The background noise of the speech data is removed by using the Wiener filtering method to obtain the speech preprocessing data.
[0089] Specifically, first, we need to obtain the statistical characteristics of the signal to be processed, including the power spectral density of the signal and the power spectral density of the noise. Then, we calculate the Wiener filter function, which is in the form of Where S x (f) is the power spectrum of the signal, S n (f) is the power spectrum of the noise. Next, the signal to be processed is converted to the frequency domain, passed through the fast Fourier transform (FFT), and multiplied by the Wiener filter function H(f). Finally, the result is converted back to the time domain through the inverse Fourier transform to obtain the denoised signal. This method can effectively reduce the impact of noise while retaining the main features of the signal.
[0090] Optionally, the preprocessing process of the posture behavior data in this embodiment includes:
[0091] 1. Data cleaning, removing noise and outliers to ensure data accuracy.
[0092] 2. Standardization: converting the data to a uniform scale for subsequent analysis.
[0093] 3. Feature extraction: extract useful features from the raw data, such as joint angle, velocity, acceleration, etc.
[0094] 4. Data segmentation: Segment continuous action data into time windows of fixed length for easy analysis.
[0095] Furthermore, the preprocessing method for text data in this embodiment includes:
[0096] 1. Word segmentation: Segment the text into individual words or phrases.
[0097] 2. Stop word removal: Remove common words that are useless for analysis, such as "de" (of), "shi" (is), etc.
[0098] 3. Lemmatization and stemming: Reduce the words to their basic forms to reduce the dimensionality of the feature space.
[0099] 4. Encoding: Convert the text into numerical form, such as bag-of-words model, TF-IDF, or use word embedding techniques (such as Word2Vec, BERT) to represent.
[0100] Preferably, extract features from the physiological preprocessing data, the facial preprocessing data, the speech preprocessing data, and the posture and behavior preprocessing data, and fuse the extracted features to obtain fused features, including:
[0101] Extract features from the physiological preprocessing data, the facial preprocessing data, the speech preprocessing data, and the posture and behavior preprocessing data respectively to obtain physiological data features, facial image features, speech data features, and posture and behavior features;
[0102] Concatenate the physiological data features, the facial image features, the speech data features, and the posture and behavior features to obtain a unified feature vector;
[0103] Perform weighted fusion on the unified feature vector to obtain the fused features.
[0104] Specifically, the extraction of physiological data features in this embodiment includes:
[0105] Time domain features: Mean, standard deviation, maximum value, minimum value, etc.
[0106] Frequency domain features: Use Fourier transform to extract spectral features, such as the main frequency and frequency energy distribution.
[0107] The extraction of facial image features includes:
[0108] Geometric features: Extract the distances and angles between facial feature points.
[0109] Expression features: Use a convolutional neural network (CNN) to extract facial expression features, or use a pre-trained model (such as VGG-Face) to extract feature vectors.
[0110] Speech data feature extraction includes
[0111] MFCC features: MFCC coefficients are extracted to represent the audio features of speech.
[0112] Emotional features: Analyze the pitch, speaking rate, and emotional tone of speech (such as fundamental frequency changes).
[0113] Furthermore, the posture phase characteristics of this embodiment include:
[0114] 1. Body position and posture:
[0115] Static posture: different body postures such as standing, sitting, lying, and the corresponding angles (such as the angle between the upper and lower body).
[0116] Dynamic posture: posture for dynamic activities such as walking, running, turning, etc.
[0117] 2. Joint angle:
[0118] The changes in the angles of various joints (such as shoulders, elbows, knees, etc.) reflect the individual's movement style and flexibility.
[0119] 3. Movement trajectory:
[0120] The path of a body in space, including dynamic characteristics such as movement speed, acceleration, and deceleration.
[0121] 4. Hand movements:
[0122] Specific hand movements (such as grasping, waving, pointing, etc.) are an important part of communication and emotional expression.
[0123] 5. Movement speed:
[0124] The speed and frequency of various activities (such as walking, running, raising hands, etc.) to reflect the individual's activity level and energy level.
[0125] 6. Symbiotic behavior:
[0126] Physical behaviors when interacting with others, such as leaning, avoiding, and guiding, reflect non-verbal information in social interactions.
[0127] 7. Posture changes:
[0128] The frequency and type of posture changes help analyze an individual's stability and adaptability.
[0129] By capturing and analyzing these gesture behavior features, this embodiment can deeply understand and evaluate the individual's emotional state, mental health, and athletic ability, etc. These features are widely used in the fields of health monitoring, emotion recognition, and human-computer interaction.
[0130] Furthermore, the voice data features of this embodiment may also include the following types of information:
[0131] 1. Emotional characteristics (emotional information):
[0132] Emotion polarity: Categorize the emotions described by the patient to identify whether they are positive, negative, or neutral.
[0133] Sentiment intensity: Quantify the intensity of sentiment expression, such as the frequency and distribution of sentiment words used.
[0134] 2. Theme characteristics (theme information):
[0135] Topic Modeling: Extracting the main topics appearing in the text through techniques such as LDA (Latent Dirichlet Allocation).
[0136] Keyword extraction: Identify the keywords and phrases most relevant to the patient’s condition.
[0137] 3. Structural features:
[0138] Text length: records the number of words or sentences in each text, which is used to analyze the level of detail of the information.
[0139] Format characteristics: the degree to which the text is structured (e.g., the use of paragraphs, subheadings, etc.).
[0140] 4. Contextual information:
[0141] Context Window: Records the context of specific words in the text to capture context-relevant information.
[0142] 5. Grammatical features:
[0143] Syntactic analysis: Analyze the structure of a sentence, including the use of noun phrases, verb phrases, etc.
[0144] Part-of-speech tagging: Marking the part of speech of each word in the text helps to understand the meaning of the sentence.
[0145] The text features in this embodiment can help doctors and researchers analyze patients' health records more effectively and assist in subsequent clinical decision-making and sentiment analysis. By extracting and analyzing these features, we can gain a deeper understanding of the patient's disease condition, treatment response, and psychological state.
[0146] Furthermore, this embodiment simply concatenates physiological features, facial features, and voice features to form a unified feature vector:
[0147] F fused =[F physiological ,F facial ,Fvoice, F attitude ]
[0148] Among them, F fused is the fused feature vector, F physiological ,F facial ,F voice, F attitude They are respectively the physiological data feature, the facial image feature, the voice data feature, and the behavioral posture feature.
[0149] Furthermore, the weighted fusion steps of this embodiment are as follows:
[0150] According to the importance of each feature to emotion classification, a weight is assigned to each feature for weighted fusion:
[0151] F fused =w1·F physiological +w2·F facial +w3F voice +w4·F attitude ;
[0152] Among them, w1, w2, w3, and w4 are the weights of each feature. Based on domain knowledge or previous research, an evaluation of the importance of each feature is given. And a preliminary weight is assigned to each feature based on literature or expert opinions.
[0153] For example, the present embodiment can also calculate the correlation between each feature and the target variable (such as the emotion classification result), select the features with higher correlation and adjust the weight accordingly. For example, the Pearson correlation coefficient or other statistical indicators can be used to evaluate the importance of the features.
[0154] Furthermore, the emotion classification module (such as support vector machine, random forest or deep learning model) is input to perform emotion classification to obtain the final detection result. The emotion classification module in this embodiment includes:
[0155] 1. Input layer, receiving fused feature vectors (such as fused features of physiological data, facial image data, voice data, and behavioral posture data). The input layer is also used to preprocess the input data (such as standardization and normalization).
[0156] 2. Feature extraction layer, which contains a series of fully connected layers, convolutional layers or recurrent layers, aims to automatically extract important features from input features. The feature extraction layer uses activation functions (such as ReLU, Sigmoid, Tanh) for nonlinear mapping.
[0157] 3. Dimensionality reduction layer, which reduces the dimensionality of the extracted features through techniques (such as principal component analysis PCA or pooling layer) to retain the most important information to reduce complexity and computational cost.
[0158] 4. Classification layer, using linear classifiers (such as logistic regression, support vector machine), neural network output layer, or ensemble methods (such as random forest, gradient boosting tree) to predict emotion categories. The output is the corresponding emotion category (such as happy, sad, angry, etc.).
[0159] 5. Loss function: Define a loss function (such as cross entropy loss) to evaluate the predictive performance of the model, targeting the gap between the classification results and the actual labels.
[0160] 6. Optimization algorithm: Use optimization algorithms (such as stochastic gradient descent SGD, Adam) to adjust model parameters and minimize the loss function.
[0161] 7. Output layer, which finally outputs the classification results and related confidence scores, which are the probability values of each emotion category.
[0162] Specifically, the model of this embodiment also includes: a feedback mechanism to implement a mechanism for adaptive adjustment of the model, and to feed back new data for model retraining during actual application to improve long-term performance.
[0163] Corresponding to the above method, this embodiment also provides an artificial intelligence-based elderly emotion detection system, including:
[0164] An artificial intelligence-based emotion detection system for the elderly, comprising:
[0165] A data acquisition module, used to acquire physiological data, facial image data, voice data, and posture behavior data of the elderly;
[0166] A preprocessing module, used to preprocess the physiological data, the facial image data, the voice data and the gesture behavior data respectively to obtain physiological preprocessing data, facial preprocessing data, voice preprocessing data and gesture behavior preprocessing data;
[0167] A fusion module, used for extracting features from the physiological preprocessing data, the facial preprocessing data, the voice preprocessing data and the posture behavior preprocessing data, and fusing the extracted features to obtain fused features;
[0168] The detection module is used to input the fusion features into the trained emotion classification module to obtain the detection results.
[0169] The beneficial effects of the present invention are as follows:
[0170] Multimodal data fusion
[0171] (1) The present invention combines physiological data, facial images, voice data and gesture behavior data to comprehensively capture the emotional state of the elderly and overcome the limitations of a single data source. Different types of data provide complementary information. For example, physiological data reflects physiological reactions, facial expressions reflect emotional states, and voice conveys emotional tone and speed.
[0172] (2) By integrating multiple data sources, the present invention significantly improves the accuracy of emotion detection and reduces the possibility of misjudgment and missed judgment. Under different environments and conditions, the system can still maintain a high recognition rate and enhance the ability to recognize complex emotions.
[0173] (3) The present invention adopts artificial intelligence technology, which can process and analyze data in real time and provide timely feedback on the emotional state of the elderly. It is suitable for scenarios that require quick response, such as smart homes and nursing systems.
[0174] (4) The present invention can learn and adjust according to the emotional characteristics of individuals, provide personalized emotional care and intervention measures, and improve the quality of life of the elderly. The present invention can be applied to various scenarios such as families, nursing homes, hospitals, etc. to meet different usage needs.
[0175] (6) The system architecture of the present invention can be continuously expanded according to demand to support the integration of more types of data input and emotion classification models.
[0176] (7) The present invention can provide decision support for family members or caregivers through the analysis of emotional data, helping them to better understand and care for the elderly. It can also conduct long-term emotional monitoring, identify emotional change trends, and provide data basis for the mental health management of the elderly.
[0177] (8) The present invention reduces dependence on human resources and reduces the cost of care and monitoring through automated emotion detection.
[0178] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0179] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of the present invention. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.
Claims
1. An artificial intelligence-based method for detecting emotions in the elderly, characterized in that: include: Obtain physiological data, facial image data, voice data, and posture behavior data of the elderly; Preprocessing the physiological data, the facial image data, the voice data and the posture behavior data respectively to obtain physiological preprocessing data, facial preprocessing data, voice preprocessing data and posture behavior preprocessing data; the voice preprocessing data is text data obtained by performing signal denoising and text conversion on the voice data; Extracting features from the physiological preprocessing data, the facial preprocessing data, the voice preprocessing data, and the posture behavior preprocessing data, and fusing the extracted features to obtain fused features; The fused features are input into the trained emotion classification module to obtain the detection results.
2. The artificial intelligence-based elderly emotion detection method according to claim 1 is characterized in that: The physiological data preprocessing step includes: translating the signal of the physiological data to obtain a translated signal; Performing wavelet decomposition on the shifted signal to obtain a plurality of wavelet coefficients; The filtering threshold is determined according to the decomposition scale and length of the signal; the calculation formula of the filtering threshold is: Among them, λ is the filtering threshold, j is the decomposition scale of the translated signal, and d j is the decomposition of wavelet coefficients with scale j, N represents the length of the signal after translation, and median represents the median operation; Constructing a denoising function according to the filtering threshold; The denoising function is used to remove the noise of the translated signal to obtain the physiological preprocessing data.
3. The method for detecting emotions of the elderly based on artificial intelligence according to claim 2 is characterized in that: Using the denoising function to remove the noise of the translated signal to obtain the physiological preprocessing data includes: The corresponding wavelet coefficients are removed by using the denoising function to obtain a smooth signal; wherein the denoising function is: Among them, X represents the wavelet coefficient, n represents the adjustment coefficient; The smoothed signal is reconstructed to obtain the physiological preprocessing data.
4. The artificial intelligence-based emotion detection method for the elderly according to claim 3 is characterized in that: The formula for signal reconstruction is: in, w j represents the energy weight of the wavelet coefficients, ||X j,k ||2 is the L2 norm of the denoised wavelet coefficients, is the reconstructed signal, J is the number of decomposition layers, ψ j',k (t) is the wavelet basis function, k is the wavelet coefficient index at a specific scale, X j,k is the kth wavelet coefficient at the jth layer.
5. The method for detecting emotions of the elderly based on artificial intelligence according to claim 1, characterized in that: The pre-processing step of the facial image data comprises: Performing filtering on the facial image data to obtain a filtered image; Constructing an image enhancement window, and using the image enhancement window to traverse the filtered image for denoising, to obtain a denoised image; The denoised image is subjected to histogram equalization processing to obtain the facial preprocessing data.
6. The method for detecting emotions of the elderly based on artificial intelligence according to claim 1, characterized in that: The pre-processing step of the speech data comprises: Using Wiener filtering method to remove background noise of the speech data; The data after background noise removal is converted into text to obtain the speech preprocessing data.
7. The method for detecting emotions of the elderly based on artificial intelligence according to claim 1, characterized in that: Extracting features from the physiological preprocessing data, the facial preprocessing data, the voice preprocessing data, and the posture behavior preprocessing data, and fusing the extracted features to obtain fused features, including: Extracting features from the physiological preprocessing data, the facial preprocessing data, the voice preprocessing data, and the posture behavior preprocessing data to obtain physiological data features, facial image features, voice data features, and posture behavior features; Performing feature splicing on the physiological data features, the facial image features, the voice data features and the posture behavior features to obtain a unified feature vector; The unified feature vector is weightedly fused to obtain the fused feature.
8. The artificial intelligence-based emotion detection method for the elderly according to claim 7 is characterized in that: The physiological data features include time domain features and frequency domain features; The facial image features include geometric features and expression features; The speech data features include MFCC features and emotion features.
9. The method for detecting emotions of the elderly based on artificial intelligence according to claim 7, characterized in that: The posture behavior characteristics include body position and posture information, joint angle information, hand movement information, movement speed information and posture change frequency information.
10. An artificial intelligence-based emotion detection system for the elderly, characterized in that: include: A data acquisition module, used to acquire physiological data, facial image data, voice data, and posture behavior data of the elderly; A preprocessing module, used to preprocess the physiological data, the facial image data, the voice data and the gesture behavior data respectively to obtain physiological preprocessing data, facial preprocessing data, voice preprocessing data and gesture behavior preprocessing data; A fusion module, used for extracting features from the physiological preprocessing data, the facial preprocessing data, the voice preprocessing data and the posture behavior preprocessing data, and fusing the extracted features to obtain fused features; The detection module is used to input the fusion features into the trained emotion classification module to obtain the detection results.
Citation Information
Patent Citations
Nuclear magnetic resonance signal acquisition and processing method and system
CN114596292A
Multi-sensory driver emotion recognition method, system and device and storage medium
CN116778461A
Elderly sentiment classification method, computer equipment and medium
CN116933129A
Nuclear magnetic resonance image enhancement method and system
CN117173038A
Cited By
Distribution box potential safety hazard detection method and system
CN119226715A
Postoperative rehabilitation nursing system for old people based on artificial intelligence
CN120376061A
Old people emotion recognition method and device based on multi-modal perception
CN120899250A
