Infant cry monitoring method, monitoring equipment, storage medium and program product

Abstract

The invention discloses a baby cry monitoring method, monitoring equipment, a storage medium and a program product, and the method comprises the steps: extracting an acoustic feature vector; establishing a corresponding relationship among the acoustic feature vectors, the physiological feature parameters and the cry types, and obtaining a feature matching rule base according to the corresponding relationship; obtaining a target acoustic feature vector; determining a first training sample infant; determining a first acoustic feature vector corresponding to the first training sample baby; calculating the acoustic similarity between the target acoustic feature vector and the first acoustic feature vector; when the acoustic similarity is greater than a preset threshold value, taking the cry type of a first acoustic feature vector corresponding to the first training sample infant as the cry type of the target infant; when the similarity is smaller than or equal to a preset threshold value, determining a second acoustic feature vector; and taking the cry type corresponding to the second acoustic feature vector as the cry type of the target infant. According to the invention, the accuracy of monitoring and identifying the cry of each baby is improved.

Description

Technical Field

[0001] This application belongs to the field of sound monitoring, and particularly relates to a method for monitoring infant crying, a monitoring device, a storage medium, and a program product. Background Art

[0002] An infant's cry is an important way for it to express needs and emotions. By the cry, it can be judged whether the infant is hungry, uncomfortable or in other states. However, since infants cannot accurately express themselves in language, parents often need to rely on experience to judge the meaning of the infant's cry. This judgment method has low accuracy, is prone to misjudging the infant's needs, and affects the physical and mental health of the infant.

[0003] Currently, some infant cry recognition technologies based on sound signal processing have emerged. By collecting the cry signals of infants, extracting sound feature parameters, and combining machine learning algorithms to classify the cry types, parents can be assisted in judging the needs of infants. By analyzing the features such as the frequency, pitch, and energy of the cry signals, this technology can more accurately identify whether the infant is crying due to hunger, sleepiness, or discomfort.

[0004] However, due to the individual differences in the cry characteristics of different infants, it is often difficult to adapt to all infants using a unified recognition model, and it is difficult to perform personalized model adjustment and optimization according to the cry characteristics of specific infants, which limits the effect of the recognition technology in practical applications and reduces the recognition accuracy. Summary of the Invention

[0005] This application provides a method for monitoring infant crying, a monitoring device, a storage medium, and a program product, which are used to improve the accuracy of monitoring and recognizing the cries of each infant.

[0006] In a first aspect, this application provides a method for monitoring infant crying, which obtains a number of cry samples of an infant training sample and corresponding label information. The label information includes the cry type and the physiological characteristic parameters of the infant training sample. The physiological characteristic parameters include head circumference, weight, and age; Extract corresponding acoustic feature vectors based on the cry samples; Establish the corresponding relationship between the acoustic feature vectors, physiological characteristic parameters, and cry types, and obtain a feature matching rule library according to the corresponding relationship; Obtain the cry signal of the target infant and the physiological characteristic parameters of the target infant; Preprocess the cry signal to obtain a target acoustic feature vector; In the feature matching rule library, determine the first training sample infant whose similarity to the physiological characteristic parameters of the target infant is greater than a preset similarity based on the physiological characteristic parameters of the target infant; Determine to take the first acoustic feature vector corresponding to the first training sample infant; Calculate the acoustic similarity between the target acoustic feature vector and the first acoustic feature vector; When the acoustic similarity is greater than a preset threshold, use the cry type of the first acoustic feature vector corresponding to the first training sample infant as the cry type of the target infant; When the similarity is less than or equal to the preset threshold, determine the second acoustic feature vector with the highest acoustic similarity to the target acoustic feature vector based on the target acoustic feature vector by means of acoustic feature clustering; Use the cry type corresponding to the second acoustic feature vector as the cry type of the target infant.

[0007] By adopting the above technical solution, a feature matching rule library is established by obtaining the cries and their label information of infant training samples, and a dual matching mechanism is adopted when identifying the cry type of the target infant. The first match is based on physiological feature parameters to screen training samples with high similarity, avoiding recognition deviations caused by using infants with large differences in weight, head circumference, and age as reference samples. When the acoustic similarity of the first match does not meet the requirements, the system will start acoustic feature clustering as a supplementary recognition means, and find the sample with the highest acoustic similarity in the feature space for matching, so as to still achieve accurate recognition when the physiological feature similarity is low. This dual matching mechanism combining physiological feature parameters and acoustic features enables the system to adapt to the sound characteristics of infants at different developmental stages, reduces recognition errors caused by individual differences of infants, and improves the accuracy and reliability of cry type recognition.

[0008] Combined with some embodiments of the first aspect, in some embodiments, establishing the correspondence relationship between the acoustic feature vector, the physiological feature parameter and the cry type specifically includes: Establish a feature space based on the acoustic feature vector; Map the physiological feature parameter to the feature space to obtain a physiological feature vector; Calculate the correlation coefficient between the acoustic feature vector and the physiological feature vector; Weight the acoustic feature vector based on the correlation coefficient to obtain a weighted acoustic feature vector; Establish the correspondence relationship between the weighted acoustic feature vector and the cry type.

[0009] By adopting the above technical solution, by mapping physiological characteristic parameters to an acoustic feature space and calculating the correlation coefficient to weight the acoustic feature vectors, the influence degree of physiological characteristics on acoustic characteristics is fully considered when establishing the feature matching rule. This weighting process enables acoustic features with strong correlation in the feature space to obtain higher weights, while the influence of features with weak correlation is appropriately suppressed. By establishing the correspondence between the weighted acoustic feature vectors and the cry types, the system can highlight the key acoustic features reflecting the infant's physiological characteristics during feature matching, reduce the interference of irrelevant features, and improve the accuracy of feature matching.

[0010] Combined with some embodiments of the first aspect, in some embodiments, calculating the correlation coefficient between the acoustic feature vector and the physiological feature vector specifically includes: Dividing the acoustic feature vector into multiple feature subsets, and each feature subset corresponds to a feature dimension; Calculating the variance contribution rate of the acoustic feature vector under each feature dimension; Selecting the feature dimensions with the cumulative variance contribution rate greater than the preset contribution rate threshold as the main feature dimensions; Calculating the grey correlation degree between the acoustic feature vector and the physiological feature vector under the main feature dimensions; Calculating the weight coefficient of each main feature dimension based on the grey correlation degree, and taking the weight coefficient as the correlation coefficient.

[0011] By adopting the above technical solution, the correlation coefficient between the acoustic feature vector and the physiological feature vector is determined by combining the variance contribution rate and the grey correlation degree. By calculating the variance contribution rate to screen the main feature dimensions, the calculation redundancy caused by too many feature dimensions is avoided. Calculating the grey correlation degree as the weight coefficient on the main feature dimensions can effectively quantify the correlation strength between the acoustic features and the physiological features. This method of calculating the correlation coefficient based on data statistics and correlation analysis enables the system to accurately identify the physiological factors that mainly affect the cry features, provides a scientific and reasonable weight basis for subsequent feature weighting, and improves the accuracy of the feature matching rule.

[0012] Combined with some embodiments of the first aspect, in some embodiments, after taking the cry type corresponding to the second acoustic feature vector as the cry type of the target infant, the method further includes: Obtaining the satisfaction time points of various physiological needs of the target infant before the current moment; Calculating the time intervals from the current moment to the last satisfaction time points of each physiological need; Establishing a quantification index of the urgency degree of physiological needs based on the time intervals; Establishing a probability correlation between the quantification index and the cry type; When there are multiple candidate cry types, prioritize the candidate cry types based on probability association to obtain a sorting result; Output the candidate cry type with the highest priority based on the sorting result as the final recognition result.

[0013] By adopting the above technical solution, a quantification index of the physiological demand urgency based on the time interval is introduced. By calculating the time interval from the current moment to the time point when each physiological demand was last satisfied, a probability association between the demand urgency and the cry type is established. When the system identifies multiple candidate cry types, the candidate types are prioritized according to the urgency of the corresponding physiological demands of each type, and the type that best matches the current physiological demand state of the baby is selected as the final recognition result. This recognition method combining time dimension information can, in the case of similar acoustic features, assist in judgment by analyzing the periodic pattern of the baby's physiological demands, effectively solving the recognition conflict problem in the traditional single feature matching method and improving the recognition accuracy of the system in complex scenarios.

[0014] Combined with some embodiments of the first aspect, in some embodiments, establishing a quantification index of the physiological demand urgency based on the time interval specifically includes: Statistically calculate the average interval period of each physiological demand of the target baby; Establish a non-linear mapping relationship between the time interval and the physiological demand urgency; Calculate the current urgency score of each physiological demand according to the non-linear mapping relationship, and use the current urgency score as the quantification index.

[0015] By adopting the above technical solution, by statistically calculating the average interval period of each physiological demand of the target baby, the system obtains the individualized physiological pattern data of the baby. Establishing a non-linear mapping relationship between the time interval and the physiological demand urgency enables the system to accurately reflect the changing characteristics of the physiological demand urgency over time, avoiding the distortion that may be caused by a simple linear mapping. The current urgency score calculated according to the non-linear mapping relationship is used as the quantification index, which not only takes into account the physiological pattern differences of individual babies but also reflects the dynamic change characteristics of the demand urgency. This non-linear quantification method based on personalized data enables the system to more accurately evaluate the urgency of different physiological demands of the baby, improves the recognition accuracy of the system in the case of multiple candidate cry types, and reduces the misjudgment probability caused by simple linear prediction.

[0016] Combined with some embodiments of the first aspect, in some embodiments, after outputting the candidate cry type with the highest priority based on the sorting result as the final recognition result, the method further includes: Record each recognition result; When the recognition result is incorrect, determine the type of reason that causes the recognition result to be incorrect; Modify the non - linear mapping relationship based on the cause type.

[0017] By adopting the above - mentioned technical solution, by recording each recognition result and determining the specific cause type that leads to errors when recognition is incorrect, the system establishes a complete error analysis mechanism. Modify the non - linear mapping relationship targeted based on the determined cause type, enabling the system to learn from errors and continuously optimize the recognition model. This adaptive correction method based on error cause analysis enables the system to continuously improve the accuracy of evaluating the urgency of infants' physiological needs. Through the dynamic adjustment of the non - linear mapping relationship, the system gradually adapts to the dynamic changes of infants' physiological laws, reduces the recognition deviation caused by the fixed mapping relationship, and improves the long - term use effect of the system.

[0018] Combined with some embodiments of the first aspect, in some embodiments, modifying the non - linear mapping relationship based on the cause type specifically includes: If the cause type is an abnormal time interval, adjust the calculation weight of the average interval period; If the cause type is an incorrect calculation of the urgency of demand, optimize the function form of the non - linear mapping relationship.

[0019] By adopting the above - mentioned technical solution, adopting a differentiated correction strategy for different error cause types, the system achieves more targeted optimization. For errors caused by abnormal time intervals, by adjusting the calculation weight of the average interval period, the system can more accurately reflect the recent changes in infants' physiological laws. For incorrect calculations of the urgency of demand, by optimizing the function form of the non - linear mapping relationship, the system improves the accuracy of evaluating the urgency of physiological needs. This differentiated correction method based on error types enables the system to precisely adjust for different types of recognition deviations while maintaining the stability of the model, improving the system's adaptive ability and recognition accuracy.

[0020] In a second aspect, an embodiment of the present application provides a monitoring device, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the system to execute the method described in the first aspect and any possible implementation manner in the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer - readable storage medium, including instructions, when the above - mentioned instructions run on the monitoring device, enabling the above - mentioned monitoring device to execute the method described in the first aspect and any possible implementation manner in the first aspect.

[0022] Fourthly, an embodiment of the present application provides a computer program product, characterized in that when the computer program product runs on a monitoring device, the monitoring device is caused to execute the method described in any possible implementation manner of the first aspect.

[0023] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. The present application provides a method for monitoring infant cries. By obtaining the cries of infant training samples and their label information, a feature matching rule library is established, and a dual matching mechanism is adopted when identifying the cry type of a target infant. The first match filters training samples with high similarity based on physiological feature parameters, avoiding recognition biases caused by using infants with large differences in weight, head circumference, and age as reference samples. When the acoustic similarity of the first match does not meet the requirements, the system will start acoustic feature clustering as a supplementary recognition means, and match by finding the sample with the highest acoustic similarity in the feature space, so as to achieve accurate recognition even when the physiological feature similarity is low. This dual matching mechanism combining physiological feature parameters and acoustic features enables the system to adapt to the sound characteristics of infants at different developmental stages, reduces recognition errors caused by individual differences in infants, and improves the accuracy and reliability of cry type recognition.

[0024] 2. The present application provides a method for monitoring infant cries. By introducing a quantification index of physiological demand urgency based on time intervals, the probability association between demand urgency and cry type is established by calculating the time interval from the current moment to the time point when each physiological demand was last satisfied. When the system identifies multiple candidate cry types, the candidate types are prioritized according to the physiological demand urgency corresponding to each type, and the type that best conforms to the current physiological demand state of the infant is selected as the final recognition result. This recognition method combining time dimension information can assist in judgment by analyzing the periodic law of infant physiological needs in the case of similar acoustic features, effectively solving the recognition conflict problem in traditional single-feature matching methods and improving the recognition accuracy of the system in complex scenarios.

[0025] 3. The present application provides a method for monitoring infant cries. By recording each recognition result and determining the specific cause type leading to errors when recognition is incorrect, the system establishes a complete error analysis mechanism. Based on the determined cause type, the non-linear mapping relationship is corrected in a targeted manner, enabling the system to learn from errors and continuously optimize the recognition model. This adaptive correction method based on error cause analysis enables the system to continuously improve the accuracy of evaluating the physiological demand urgency of infants. By dynamically adjusting the non-linear mapping relationship, the system gradually adapts to the dynamic changes of infant physiological laws, reduces recognition biases caused by fixed mapping relationships, and improves the long-term use effect of the system. Description of the Drawings

[0026] Figure 1 It is a schematic flowchart of a method for monitoring infant crying in an embodiment of the present application.

[0027] Figure 2 It is a schematic flowchart of an optimization method based on the time sequence law of physiological needs in an embodiment of the present application.

[0028] Figure 3 It is a schematic structural diagram of an entity device of an infant crying monitoring device provided in an embodiment of the present application. Detailed implementation manners

[0029] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification and appended claims of the present application, the singular forms "a", "an", "the", "above-mentioned", "said", and "this" are intended to include the plural forms as well, unless the context clearly dictates otherwise. It should also be understood that the term "and / or" used in the present application refers to any and all possible combinations of one or more of the listed items.

[0030] Hereinafter, the terms "first" and "second" are only used for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0031] Next, an embodiment is used in combination with Figure 1 to describe a method for monitoring infant crying in an embodiment of the present application: Please refer to Figure 1 which is a schematic flowchart of a method for monitoring infant crying in an embodiment of the present application.

[0032] S101. Obtain a number of crying samples of the infant training sample and the corresponding label information, and extract the corresponding acoustic feature vectors based on the crying samples; The monitoring device obtains a number of crying samples of the infant training sample and the corresponding label information. The label information includes the crying type and the physiological characteristic parameters of the infant training sample. The physiological characteristic parameters include head circumference, weight, and age. Then, the corresponding acoustic feature vectors are extracted based on the crying samples.

[0033] In this step, the monitoring device needs to obtain a number of cry samples of the infant training samples and the corresponding label information. The cry samples can be pre-recorded infant crying audio data or real-time collected infant crying audio data. The label information is the information corresponding to the cry samples, including the cry type and the physiological characteristic parameters of the infant training samples. The physiological characteristic parameters can include, but are not limited to, head circumference, weight, age, etc. The monitoring device can obtain these cry samples and label information through various means, such as reading from a database, receiving from other devices, or manually input by the user, etc.

[0034] After obtaining the cry samples, the monitoring device needs to extract the corresponding acoustic feature vectors based on the cry samples. The acoustic feature vectors are a set of numerical values obtained by extracting features from the audio data and can represent the characteristics of the audio data. The monitoring device can adopt various audio feature extraction algorithms to extract the acoustic feature vectors, such as Mel Frequency Cepstral Coefficients (MFCC), Linear Predictive Coding (LPC), Perceptual Linear Prediction (PLP), etc. The extracted acoustic feature vectors are usually a multi-dimensional vector, and each dimension represents a feature value.

[0035] It should be noted that when extracting the acoustic feature vectors, some problems may be encountered, such as environmental noise interference, incomplete cry signals, etc. To solve these problems, the monitoring device can preprocess the cry samples, such as noise reduction, endpoint detection, etc., to improve the quality of the acoustic feature vectors. In addition, the monitoring device can also adopt some data augmentation techniques, such as adding random noise, changing the pitch, etc., to increase the diversity of the training samples and improve the generalization ability of the model.

[0036] S102. Establish the correspondence relationship among the acoustic feature vectors, physiological characteristic parameters and cry types, and obtain the feature matching rule library according to the correspondence relationship; The monitoring device establishes the correspondence relationship among the acoustic feature vectors, physiological characteristic parameters and cry types. Specifically: establish a feature space based on the acoustic feature vectors; map the physiological characteristic parameters to the feature space to obtain the physiological feature vectors; calculate the correlation coefficient between the acoustic feature vectors and the physiological feature vectors. Specifically: divide the acoustic feature vectors into multiple feature subsets, and each feature subset corresponds to a feature dimension; calculate the variance contribution rate of the acoustic feature vectors under each feature dimension; select the feature dimensions with the cumulative variance contribution rate greater than the preset contribution rate threshold as the main feature dimensions; calculate the grey correlation degree between the acoustic feature vectors and the physiological feature vectors under the main feature dimensions; calculate the weight coefficients of each main feature dimension based on the grey correlation degree and use the weight coefficients as the correlation coefficients; weight the acoustic feature vectors based on the correlation coefficients to obtain the weighted acoustic feature vectors; establish the correspondence relationship between the weighted acoustic feature vectors and the cry types. After that, the monitoring device obtains the feature matching rule library according to the correspondence relationship.

[0037] In this step, the monitoring device needs to establish the correspondence between the acoustic feature vectors, physiological characteristic parameters, and cry types. Specifically, the monitoring device first establishes a feature space based on the acoustic feature vectors, then maps the physiological characteristic parameters into the feature space to obtain physiological feature vectors. Next, the monitoring device calculates the correlation coefficient between the acoustic feature vectors and the physiological feature vectors. The specific steps include: dividing the acoustic feature vectors into multiple feature subsets, with each feature subset corresponding to a feature dimension; calculating the variance contribution rate of the acoustic feature vectors under each feature dimension; selecting the feature dimensions with the cumulative variance contribution rate greater than the preset contribution rate threshold as the main feature dimensions; calculating the grey correlation degree between the acoustic feature vectors and the physiological feature vectors under the main feature dimensions; calculating the weight coefficient of each main feature dimension based on the grey correlation degree, and using the weight coefficient as the correlation coefficient. Finally, the monitoring device weights the acoustic feature vectors based on the correlation coefficient to obtain weighted acoustic feature vectors, and establishes the correspondence between the weighted acoustic feature vectors and the cry types.

[0038] When establishing the correspondence, the monitoring device can adopt various machine learning algorithms, such as support vector machine (SVM), random forest (Random Forest), artificial neural network (ANN), etc. These algorithms can learn the mapping relationship between the acoustic feature vectors, physiological characteristic parameters, and cry types through training, so as to establish a classification model. The monitoring device can select a suitable algorithm according to the specific situation and adjust the parameters of the algorithm to obtain the best performance.

[0039] It should be noted that when establishing the correspondence, some problems may be encountered, such as sample imbalance, feature redundancy, etc. To solve these problems, the monitoring device can adopt some technical means, such as oversampling, undersampling, feature selection, etc., to optimize the training data and feature space. In addition, the monitoring device can also introduce some prior knowledge, such as the crying characteristics of infants of different ages, the acoustic features of different cry types, etc., to guide the training and optimization of the model. Finally, the monitoring device can store the established correspondence in the feature matching rule library for subsequent use. The feature matching rule library can adopt various data structures for organization and storage, such as hash tables, decision trees, etc., to achieve efficient query and matching.

[0040] S103. Obtain the cry signal of the target infant and the physiological characteristic parameters of the target infant, and preprocess the cry signal to obtain the target acoustic feature vector; In this step, the monitoring device needs to obtain the cry signal of the target infant and the physiological characteristic parameters of the target infant. The cry signal can be collected in real time through audio collection devices such as microphones, or read from pre-recorded audio files. The physiological characteristic parameters can be input manually or obtained from other devices, such as weighing scales, rulers, etc.

[0041] After obtaining the cry signal, the monitoring device needs to preprocess it to obtain the target acoustic feature vector. The preprocessing may include but is not limited to the following steps: signal framing, windowing, endpoint detection, noise reduction, pre-emphasis, etc. These preprocessing steps can improve the quality of the signal, reduce the influence of noise and interference, and provide better input for subsequent feature extraction.

[0042] The monitoring device can use the same feature extraction algorithm as in step S101, such as MFCC, LPC, PLP, etc., to extract the target acoustic feature vector. The extracted target acoustic feature vector has the same dimension and physical meaning as the acoustic feature vector in step S101 and can be used for subsequent feature matching and classification.

[0043] It should be noted that in practical applications, the cry signal of the target infant may be affected by various factors, such as environmental noise, differences in acquisition devices, etc. To improve the robustness of the system, the monitoring device can adopt some adaptive algorithms, such as adaptive noise cancellation, vocal tract length normalization, etc., to dynamically adjust the parameters of preprocessing and feature extraction to adapt to different application scenarios. In addition, the monitoring device can also use multiple microphone arrays to achieve sound source localization and separation, so as to obtain a purer cry signal and improve the accuracy of feature extraction.

[0044] S104. In the feature matching rule library, determine the first training sample infant whose similarity to the physiological feature parameters of the target infant is greater than the preset similarity based on the physiological feature parameters of the target infant; In this step, the monitoring device needs to determine the first training sample infant whose similarity to the physiological feature parameters of the target infant is greater than the preset similarity in the feature matching rule library. Specifically, the monitoring device first reads the physiological feature parameters of all training sample infants from the feature matching rule library, and then calculates the similarity between the physiological feature parameters of each training sample infant and the physiological feature parameters of the target infant. The similarity can be measured in various ways, such as Euclidean distance, cosine similarity, Pearson correlation coefficient, etc.

[0045] After calculating the similarity between the physiological feature parameters of all training sample infants and the physiological feature parameters of the target infant, the monitoring device can set a preset similarity threshold, such as 0.8 or 0.9, and then select all training sample infants with similarity greater than this threshold as the candidate set. If there is only one training sample infant in the candidate set, then this infant is the first training sample infant; if there are multiple training sample infants in the candidate set, then the monitoring device can select the one with the greatest similarity as the first training sample infant.

[0046] It should be noted that the selection of the preset similarity threshold will affect the subsequent matching accuracy and efficiency. If the threshold is set too high, there may be no training sample infants that meet the conditions, resulting in the inability to perform subsequent matching; if the threshold is set too low, there may be too many training sample infants in the candidate set, thus reducing the matching efficiency. Therefore, the monitoring device can select the optimal preset similarity threshold through cross-validation and other methods according to the actual situation to achieve a balance between accuracy and efficiency.

[0047] S105. Determine the first acoustic feature vector corresponding to the first training sample infant; In this step, the monitoring device needs to determine the first acoustic feature vector corresponding to the first training sample infant. The first training sample infant is the training sample infant that is most similar to the target infant determined based on the physiological feature parameter similarity in step S104. Each training sample infant has a corresponding acoustic feature vector in the feature matching rule library. Therefore, the monitoring device only needs to read the acoustic feature vector corresponding to the first training sample infant from the feature matching rule library.

[0048] It should be noted that the first acoustic feature vector and the target acoustic feature vector have the same dimension and physical meaning. They are both a set of values obtained by extracting features from the infant crying signal and can be used for subsequent feature matching and classification. Therefore, when the monitoring device reads the first acoustic feature vector, it needs to ensure that its format and content are consistent with the target acoustic feature vector for subsequent processing.

[0049] S106. Calculate the acoustic similarity between the target acoustic feature vector and the first acoustic feature vector; In this step, the monitoring device needs to calculate the acoustic similarity between the target acoustic feature vector and the first acoustic feature vector. Acoustic similarity represents the similarity degree between two acoustic feature vectors and can be used to measure the similarity between the crying signal of the target infant and the crying signal of the first training sample infant.

[0050] The monitoring device can adopt various similarity measurement methods to calculate the acoustic similarity, such as Euclidean distance, cosine similarity, KL divergence, etc. These methods are all based on the vector space model, regarding the acoustic feature vector as a point in a high-dimensional space, and then calculating the distance or angle between two points to represent their similarity degree. For example, the Euclidean distance calculates the straight-line distance between two vectors, and the smaller the distance, the more similar; the cosine similarity calculates the cosine value of the angle between two vectors, and the smaller the angle, the more similar.

[0051] When selecting a similarity measurement method, the monitoring device needs to consider the characteristics and distribution of the acoustic feature vectors. For example, if the acoustic feature vector is a high-dimensional sparse vector, the Euclidean distance may not be very suitable because it is affected by the curse of dimensionality; if the acoustic feature vector is a normalized unit vector, the cosine similarity may be a better choice because it is insensitive to the vector length. Therefore, the monitoring device can select the optimal similarity measurement method through experimental comparison and other means according to the actual situation to achieve the best matching effect.

[0052] It should be noted that when calculating the acoustic similarity, some problems may be encountered, such as inconsistent feature vector dimensions, large differences in numerical ranges, etc. To solve these problems, the monitoring device can preprocess the acoustic feature vectors, such as normalization, standardization, dimensionality reduction, etc., to eliminate the dimensional and numerical differences between different features. In addition, the monitoring device can also introduce some prior knowledge, such as the acoustic feature distributions of different cry types, to guide the calculation of similarity and the selection of thresholds to improve the accuracy and reliability of matching.

[0053] S107. When the acoustic similarity is greater than the preset threshold, use the cry type of the first acoustic feature vector corresponding to the first training sample baby as the cry type of the target baby; This step is to judge the acoustic similarity between the target acoustic feature vector and the first acoustic feature vector calculated previously. When the acoustic similarity is greater than the preset threshold, it indicates that the cry signal of the target baby has a high similarity with the cry signal of the first training sample baby, and it can be considered that the cry type of the target baby is the same as that of the first training sample baby. Therefore, the monitoring device can directly determine the cry type of the first acoustic feature vector corresponding to the first training sample baby as the cry type of the target baby.

[0054] In specific implementation, the monitoring device can set an acoustic similarity threshold, such as 0.8 or 0.9. When the calculated acoustic similarity is higher than this threshold, it is considered that the cry type of the target baby matches the cry type of the first training sample baby. The acoustic similarity threshold can be set according to actual needs. The higher the threshold, the higher the accuracy of the judgment, but the number of matching samples may decrease; the lower the threshold, although the number of matching samples will increase, the probability of false matching will also increase. Therefore, it is necessary to balance the accuracy and the number of matching samples and set an appropriate threshold.

[0055] If the acoustic similarity is not greater than the preset threshold, it indicates that there is a significant difference between the cry signal of the target infant and the cry signal of the first training sample infant, and it cannot be simply considered that they belong to the same cry type. In this case, other judgment methods may need to be considered, such as increasing the number of training samples to improve the flexibility of matching; or considering using other acoustic feature extraction methods to obtain more discriminative acoustic feature vectors and improve the accuracy of matching.

[0056] S108. When the similarity is less than or equal to the preset threshold, determine the second acoustic feature vector with the highest acoustic similarity to the target acoustic feature vector by using acoustic feature clustering based on the target acoustic feature vector; This step is to adopt another method to determine the cry type of the target infant when the acoustic similarity does not meet the preset threshold. Since there is a significant difference between the cry signal of the first training sample infant and the cry signal of the target infant, the cry type of the first training sample infant cannot be directly used, and it is necessary to find samples closer to the cry signal of the target infant in a larger sample space.

[0057] The monitoring device can use acoustic feature clustering to cluster the acoustic feature vectors of all training samples, gather similar acoustic feature vectors together to form multiple clustering clusters. Then, the monitoring device can calculate the similarity between the target acoustic feature vector and the center of each clustering cluster, find the clustering cluster with the highest similarity to the target acoustic feature vector, and determine the acoustic feature vector with the highest similarity to the target acoustic feature vector within this clustering cluster as the second acoustic feature vector.

[0058] Various clustering algorithms can be used for acoustic feature clustering, such as K-means clustering, hierarchical clustering, DBSCAN clustering, etc. Different clustering algorithms are suitable for different data distribution situations, and the monitoring device can select a suitable clustering algorithm according to the actual distribution characteristics of the acoustic feature vectors. At the same time, the clustering effect is also related to the number of clustering clusters. Too many or too few clustering clusters will affect the clustering quality, and it is necessary to set a suitable number of clustering clusters according to the number and distribution of samples.

[0059] When looking for the second acoustic feature vector, if there is only one sample inside the most similar clustering cluster, the acoustic feature vector corresponding to this sample is the second acoustic feature vector; if there are multiple samples inside the most similar clustering cluster, it is necessary to further calculate the similarity between the target acoustic feature vector and each sample, and the acoustic feature vector corresponding to the sample with the highest similarity is the second acoustic feature vector.

[0060] S109. Take the cry type corresponding to the second acoustic feature vector as the cry type of the target infant.

[0061] This step is to determine the cry type of the target infant based on the second acoustic feature vector found in the previous step. Since the second acoustic feature vector is the acoustic feature vector that is most similar to the target acoustic feature vector found in a larger sample space through clustering, the cry type corresponding to the second acoustic feature vector can be used as the cry type of the target infant.

[0062] The monitoring device can directly extract the cry type label corresponding to the second acoustic feature vector as the cry type of the target infant. Since the second acoustic feature vector is obtained through clustering and has a high similarity to the target acoustic feature vector, the corresponding cry type is also more likely to match the actual cry type of the target infant.

[0063] Of course, since the clustering result is affected by the clustering algorithm and clustering parameters, it cannot be completely guaranteed that the second acoustic feature vector must correspond to the correct cry type. To improve the accuracy of judgment, the monitoring device can further analyze and verify the second acoustic feature vector, such as calculating the difference degree between the second acoustic feature vector and the target acoustic feature vector to determine whether the difference degree is within an acceptable range; or calculating the confidence level of the cry type corresponding to the second acoustic feature vector, and only when the confidence level is high enough is the judgment result considered credible.

[0064] If the credibility of the second acoustic feature vector is not high enough, it may be necessary to consider introducing other judgment methods, such as making a comprehensive judgment by combining other physiological characteristic parameters of the infant, or having professional medical staff conduct auditory analysis on the cry signal. In short, when determining the cry type of the target infant, it is necessary to improve the accuracy of judgment as much as possible to avoid the impact of incorrect judgment results on subsequent processing and decision-making.

[0065] In the above embodiment, a feature matching rule library is established by obtaining the cries and their label information of infant training samples, and a dual matching mechanism is adopted when identifying the cry type of the target infant. The first match is based on physiological characteristic parameters to screen training samples with high similarity, avoiding recognition deviation caused by using infants with large differences in weight, head circumference, and age as reference samples. When the acoustic similarity of the first match does not meet the requirements, the system will start acoustic feature clustering as a supplementary recognition means to find the sample with the highest acoustic similarity in the feature space for matching, so as to achieve accurate recognition even when the physiological feature similarity is low. This dual matching mechanism combining physiological characteristic parameters and acoustic features enables the system to adapt to the voice characteristics of infants at different developmental stages, reduces recognition errors caused by individual differences of infants, and improves the accuracy and reliability of cry type recognition.

[0066] The above embodiment realizes the basic recognition of the infant crying type through the dual matching mechanism of physiological characteristic parameters and acoustic characteristics. However, in actual application, there may be multiple candidate crying types, and it is difficult to accurately judge the final recognition result only by relying on the similarity of acoustic characteristics. Therefore, the present application also provides an optimization method based on the temporal law of physiological needs. By analyzing the time law of satisfying various physiological needs of infants, a probabilistic association between the urgency of needs and the crying type is established, so that a more accurate judgment can be made when multiple candidate types appear. The following is combined with Figure 2 , an optimization method based on the physiological demand timing law in an embodiment of the present application is described: See also Figure 2 , which is a flow chart of an optimization method based on the timing law of physiological needs in an embodiment of the present application.

[0067] S201, obtaining the time points at which multiple physiological needs of the target infant are satisfied before the current moment; This step collects data on the historical physiological needs of the target infant. An infant's physiological needs primarily include eating, sleeping, and elimination, and these needs follow certain cyclical patterns. By obtaining the time points at which the target infant's various physiological needs were met before the current moment, we can understand the regularity of their physiological needs, providing a data foundation for subsequent analysis and judgment.

[0068] Monitoring devices can capture historical physiological needs in a variety of ways. For example, they can connect to an infant monitoring system to automatically record the times of events like eating, sleeping, and defecating. Alternatively, they can rely on manual input from parents, allowing them to record the times each time their infant's physiological needs were met. Regardless of the method used, monitoring devices must perform necessary data cleaning and processing to eliminate abnormal or erroneous data and ensure data accuracy and completeness.

[0069] When acquiring historical data, the monitoring device can set an appropriate time window based on actual needs, such as acquiring data from the past week or month. The size of the time window requires a balance between data volume and timeliness. A time window that is too large will result in excessive data volume, increasing the computational and storage burdens; a time window that is too small may not fully reflect the regularity of the infant's physiological needs. Furthermore, the monitoring device can set different time windows for different types of physiological needs. For example, the time window for dietary needs can be set shorter, while the time window for sleep needs can be set longer.

[0070] S202, calculating the time interval between the current moment and the last time each physiological need was satisfied; This step is to calculate the time intervals between the current moment and the last satisfaction time points of various physiological needs based on the acquired historical data. The time intervals can reflect how long it has been since the last satisfaction of the physiological needs at the current moment, and indirectly indicate the urgency of the current physiological needs.

[0071] The monitoring device can find the time point of the most recent satisfaction of each physiological need, and then calculate the time difference between the current moment and this time point to obtain the time interval. The time interval can be expressed in various time units, such as seconds, minutes, hours, etc., and the monitoring device can select an appropriate time unit according to actual needs.

[0072] When calculating the time interval, the monitoring device needs to pay attention to handling some special situations. For example, if a long time has passed since the last satisfaction time point at the current moment, which has exceeded the average occurrence period of the physiological need, this may mean that one or more opportunities to satisfy the need have been missed, and the calculation method of the time interval needs to be adjusted accordingly. In addition, if there is no satisfaction record for a certain physiological need in the historical data, a default time interval value can be considered, or the physiological need can be directly ignored.

[0073] S203. Establish a quantitative index for the urgency of physiological needs based on the time interval; The monitoring device establishes a quantitative index for the urgency of physiological needs based on the time interval, which specifically includes: Statistical average interval period of each physiological need of the target infant; Establish a non-linear mapping relationship between the time interval and the urgency of physiological needs; Calculate the current urgency score of each physiological need according to the non-linear mapping relationship, and use the current urgency score as the quantitative index.

[0074] This step is to convert the time interval into a quantitative index for the urgency of physiological needs, and use a numerical value to represent the urgency of the current physiological needs. The urgency index can provide a basis for establishing the probability association of cry types in the future.

[0075] The monitoring device establishes a quantitative index for the urgency of physiological needs based on the time interval, and the following methods can be specifically adopted: First, statistical average interval period of each physiological need of the target infant. The average interval period reflects how often an infant usually has a certain physiological need, and can be estimated by the average time difference between adjacent satisfaction time points in the historical data.

[0076] Then, establish a non - linear mapping relationship between the time interval and the physiological need urgency. The urgency and the time interval do not have a simple linear relationship. The longer the time interval, the faster the urgency may increase. Therefore, a non - linear mapping function needs to be established to map the time interval to the urgency value. Common non - linear functions can include exponential functions, logarithmic functions, power functions, etc.

[0077] Finally, according to the non - linear mapping relationship, calculate the current urgency scores of each physiological need, and use the current urgency scores as quantization indicators. Substitute the time interval at the current moment into the non - linear mapping function, and the current urgency score can be obtained. The higher the score, the more urgent the physiological need.

[0078] S204. Establish a probability association between the quantization indicator and the cry type; This step is to establish a probability association relationship between physiological needs and cry types based on the physiological need urgency quantization indicator. Different physiological needs may cause infants to emit different types of cries. By analyzing the corresponding relationship between the satisfaction of physiological needs and cry types in historical data, the association probability between the two can be obtained.

[0079] The monitoring device can use data mining or machine learning methods to automatically learn the association rules between physiological needs and cry types from historical data. Specifically, pair the time points when each physiological need is satisfied with the adjacent cry occurrence time points to form a sample data of physiological need - cry type. Then, use association rule mining algorithms, such as the Apriori algorithm or the FP - Growth algorithm, to mine the frequent association patterns between physiological needs and cry types from the sample data, and obtain the association rules and the corresponding confidence levels. The confidence level can be used as a probability measure of the association strength between the two.

[0080] When establishing the probability association, the monitoring device can consider the mutual influence and combined effects between different physiological needs. Some physiological needs may occur simultaneously, such as hunger and thirst, and they may produce a combined cry type. To more accurately characterize this combined effect, the monitoring device can use a multi - dimensional association rule mining algorithm to combine the urgency indicators of multiple physiological needs into a composite feature, and then establish an association with the cry type.

[0081] In practical applications, there may be situations where the association rules between certain physiological needs and cry types are too complex or uncertain. For example, due to insufficient sample data or noise interference, the confidence of the mined association rules is low or there are multiple conflicting rules. In such cases, the monitoring device can introduce domain knowledge or expert experience to manually evaluate and screen the mined association rules, eliminating rules with too low confidence or unreasonable ones. At the same time, the association rules can be updated and optimized regularly, adjusting and supplementing the rules according to the newly collected sample data to improve the accuracy of the association probability.

[0082] S205. When there are multiple candidate cry types, prioritize the candidate cry types based on probability association to obtain a ranking result. This step is to prioritize the candidate cry types using the probability association between physiological needs and cry types after identifying multiple possible cry types. The prioritization can provide a reference for the final determination of the cry type.

[0083] The monitoring device can calculate the association probability between each candidate cry type and the current urgency of physiological needs as the priority score for this candidate type. The association probability can be calculated and predicted based on the probability association model established in step S204. Specifically, the following two methods can be used: One is the prediction method based on association rules. The monitoring device finds the association rules satisfied by the current physiological need urgency index, and then calculates the association probability score for each candidate cry type according to the confidence of the rules. If multiple rules are satisfied simultaneously, the rule with the highest confidence is taken as the prediction basis.

[0084] The other is the prediction method based on a probabilistic graphical model. The monitoring device constructs a probabilistic graphical model, such as a Bayesian network or a Markov random field, with the physiological need urgency index as the input node, the candidate cry types as the output nodes, and the intermediate nodes representing the association probability between the two. Then, a probabilistic inference algorithm is used to calculate the posterior probability of each candidate cry type as the association probability score.

[0085] After calculating the association probability scores for all candidate cry types, the monitoring device sorts the candidate types in descending order according to the scores to obtain the prioritization result. The candidate types ranked higher indicate a higher degree of association with the current physiological needs and are more likely to be the actual cause of the baby's cry.

[0086] When sorting, it may be encountered that the correlation probability scores of multiple candidate types are very close, resulting in unstable sorting results. For such a situation, the monitoring device can set a score difference threshold. When the score difference between two consecutive candidate types is less than the threshold, it is considered that their priorities are basically the same and they can be sorted side by side. In addition, the monitoring device can also introduce other sorting bases, such as the prior probability and historical frequency of the candidate type, and perform weighted combination with the correlation probability score to obtain a more comprehensive sorting result.

[0087] S206. Output the candidate cry type with the highest priority as the final recognition result based on the sorting result.

[0088] This step is to output the candidate cry type with the highest confidence as the final recognition result of the baby's cry according to the priority sorting result. Through the optimization of the physiological demand time sequence law, the most likely reason can be selected from multiple candidate types, improving the accuracy of cry recognition.

[0089] The monitoring device can directly take the first place in the priority sorting as the final recognition result for output. At the same time, the confidence of the recognition result can also be output, indicating the credibility of this result. The confidence can be measured by the difference between the priority score and the scores of other candidate types. The greater the difference, the higher the confidence.

[0090] When outputting the recognition result, the monitoring device can adopt various forms, such as text prompts, voice announcements, graphical interface displays, etc., to facilitate parents or caregivers to timely understand the reason for the baby's cry and take corresponding measures to meet the baby's needs.

[0091] Considering that there is still a certain degree of uncertainty in the recognition result, the monitoring device can simultaneously output other candidate types with relatively high possibilities as alternative results for reference. These alternative results can be listed in sequence according to the priority sorting, or a confidence threshold can be set to list all candidate types with a confidence higher than the threshold.

[0092] In practical applications, after the monitoring device obtains the final recognition result, it can also further verify and correct the result in combination with other information such as environmental factors and the baby's state. For example, if the recognition result is hunger, but the monitoring data shows that the baby has just eaten not long ago, then the accuracy of the recognition result may need to be suspected and other possible reasons should be considered. Another example is that if the recognition result is that the diaper is wet, but the humidity sensor shows that the diaper is dry, then it may be necessary to check whether the sensor fails or re-evaluate the recognition model. In short, when outputting the final recognition result, various possible influencing factors need to be fully considered to improve the accuracy and reliability of recognition as much as possible.

[0093] In the above embodiments, a quantification index of physiological demand urgency based on time interval is introduced. By calculating the time interval between the current moment and the time point when each physiological demand was last satisfied, a probability correlation between demand urgency and cry types is established. When the system identifies multiple candidate cry types, the candidate types are prioritized according to the urgency of the corresponding physiological demands, and the type that best matches the current physiological demand state of the infant is selected as the final recognition result. This recognition method combining time dimension information can, in the case of similar acoustic features, assist in judgment by analyzing the periodic pattern of the infant's physiological demands, effectively solving the recognition conflict problem in traditional single-feature matching methods and improving the recognition accuracy of the system in complex scenarios.

[0094] Further, in another embodiment, after outputting the candidate cry type with the highest priority as the final recognition result based on the sorting result, the method further includes: Recording each recognition result; When the recognition result is incorrect, determining the type of reason that causes the recognition result to be incorrect; Based on the type of reason, correcting the non-linear mapping relationship, specifically including: If the type of reason is an abnormal time interval, adjusting the calculation weight of the average interval period; If the type of reason is an incorrect calculation of demand urgency, optimizing the functional form of the non-linear mapping relationship.

[0095] This embodiment provides a feedback optimization mechanism for recognition results. By recording and analyzing the accuracy of each recognition result, finding out the reasons for recognition errors, and making targeted corrections and optimizations to the recognition model for different types of reasons, the recognition accuracy and reliability are continuously improved. The steps of this embodiment are described in detail below: Recording each recognition result: This step records and stores the results of each cry recognition, providing a data basis for subsequent analysis and optimization. The monitoring device can compare the recognition result with the actual situation, mark the correct and incorrect recognition results, and store them in the recognition result database. The recognition result database can contain multiple fields, such as recognition time, candidate type, priority score, final result, correctness mark, etc., facilitating subsequent query and analysis.

[0096] When recording the recognition result, the monitoring device can also synchronously record other relevant information, such as the physiological state of the infant, environmental parameters, etc., for analyzing the reasons for recognition errors. To ensure the integrity and consistency of the data, the monitoring device can adopt a transaction mechanism to submit the recognition result and other information as an atomic transaction, avoiding problems such as partial data loss or out-of-sync.

[0097] When the recognition result is incorrect, determine the type of reason that causes the recognition result to be incorrect: This step is to analyze the reasons for the incorrect recognition result and find out the key factors leading to the recognition error. The monitoring device can start from multiple aspects and comprehensively analyze the reasons for the incorrect recognition: First, analyze whether the time interval of the physiological needs is abnormal. By comparing the time interval at the time of incorrect recognition with the average interval period, judge whether there is an abnormal deviation. If the time interval significantly exceeds the normal range, it may lead to inaccurate urgency estimation, which in turn affects the recognition result.

[0098] Second, analyze whether the calculation of the physiological need urgency is accurate. By comparing the urgency score at the time of incorrect recognition with the actual need situation, judge whether there is a deviation in the non-linear mapping relationship. If the mapping function cannot well reflect the relationship between the time interval and the urgency, it may underestimate or overestimate the urgency of some needs.

[0099] Third, analyze whether there are errors in the cry feature extraction and classification model. By comparing the cry features at the time of incorrect recognition with the standard features, as well as the confidence of the classification model, judge whether there are deviations in feature extraction or classification decision-making. If the selected features cannot well describe the differences between different cry types, or the classification model is overfitted or underfitted, it may lead to recognition errors.

[0100] Fourth, analyze whether environmental factors and the baby's state interfere with the recognition. By analyzing the environmental parameters and the baby's physiological indicators at the time of recognition, judge whether there are problems such as unclear audio collection and abnormal baby state. Factors such as environmental noise and echo may affect the quality of the cry signal, and special states such as the baby being sick or excited may also cause changes in cry features.

[0101] The monitoring device comprehensively analyzes the above factors and judges the main type of reason that causes the recognition error by setting a series of rules and thresholds. For example, if the time interval deviates from the average value by more than 50% and other factors are relatively normal, the reason type is determined to be abnormal time interval; if the difference between the urgency and the actual need exceeds 2 standard deviations and other factors are relatively normal, the reason type is determined to be incorrect urgency calculation. Of course, in actual applications, there may be cases where multiple factors act together to cause recognition errors. In this regard, the monitoring device can further refine the reason type or set a measurement index for the comprehensive influence degree to find the one or several reasons with the greatest influence.

[0102] In the above embodiments, by recording each recognition result and determining the specific cause type that leads to errors when recognition fails, the system establishes a complete error analysis mechanism. Based on the determined cause type, targeted correction is made to the non-linear mapping relationship, enabling the system to learn from errors and continuously optimize the recognition model. This adaptive correction method based on error cause analysis enables the system to continuously improve the accuracy of assessing the urgency of infants' physiological needs. By dynamically adjusting the non-linear mapping relationship, the system gradually adapts to the dynamic changes in infants' physiological laws, reduces the recognition deviation caused by the fixed mapping relationship, and improves the long-term usage effect of the system.

[0103] The monitoring device in the embodiments of the present invention application will be described from the perspective of hardware processing. Please refer to Figure 3 , which is a schematic structural diagram of an entity device of a baby cry monitoring device provided in the embodiments of the present application.

[0104] It should be noted that Figure 3 the structure of the monitoring device shown is only an example and should not impose any limitations on the functions and usage scope of the embodiments of the present invention.

[0105] As Figure 3 shown, the monitoring device includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the Read-Only Memory (ROM) 302 or the program loaded from the storage section 308 into the Random Access Memory (RAM) 303, such as executing the method in the above embodiments. In the RAM 303, various programs and data required for system operation are also stored. The CPU 301, ROM 302, and RAM 303 are connected to each other through a bus 304. The Input / Output (I / O) interface 305 is also connected to the bus 304.

[0106] The following components are connected to the I / O interface 305: an input section 306 including a camera, an infrared sensor, etc.; an output section 307 including a liquid crystal display (LCD), a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0107] Specifically, according to an embodiment of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, an embodiment of the present invention includes a computer program product including a computer program carried on a computer-readable medium, the computer program including a computer program for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 309, and / or installed from the removable medium 311. When the computer program is executed by a central processing unit (CPU) 301, various functions defined in the present invention are executed.

[0108] It should be noted that the computer-readable medium shown in the embodiments of the present invention can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device. In the present invention, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable computer program. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above.

[0109] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present invention. Among them, each block in the flowchart or block diagram can represent a module, a program segment, or a part of code, and the above module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, as well as the combination of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0110] As another aspect, the present invention also provides a computer-readable storage medium, which may be included in the system described in the above embodiments; or may exist alone without being assembled into the system. The above storage medium carries one or more computer programs, and when the one or more computer programs are executed by a processor of a system, the system implements the method provided in the above embodiments.

[0111] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

[0112] As used in the above embodiments, depending on the context, the term "when..." can be interpreted to mean "if...", or "after...", or "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if detecting (the stated condition or event)" can be interpreted to mean "if determining...", or "in response to determining...", or "when detecting (the stated condition or event)", or "in response to detecting (the stated condition or event)".

[0113] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer may be a general-purpose computer, a dedicated computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid-state drive), etc.

[0114] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by relevant hardware instructed by a computer program. This program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. The foregoing storage medium includes various media that can store program codes, such as ROM or random access memory RAM, magnetic disks, or optical discs.

Claims

1. A method for monitoring infant crying, characterized in that, Including: Obtaining a number of cry samples of the infant training samples and corresponding label information, where the label information includes the cry type and the physiological characteristic parameters of the infant training samples, and the physiological characteristic parameters include head circumference, weight, and age; Extracting corresponding acoustic feature vectors based on the cry samples; Establishing the correspondence between the acoustic feature vectors, the physiological characteristic parameters, and the cry type, and obtaining a feature matching rule library according to the correspondence; Obtaining the cry signal of the target infant and the physiological characteristic parameters of the target infant; Preprocessing the cry signal to obtain a target acoustic feature vector; In the feature matching rule library, determining a first training sample infant whose similarity to the physiological characteristic parameters of the target infant is greater than a preset similarity based on the physiological characteristic parameters of the target infant; Determining to obtain the first acoustic feature vector corresponding to the first training sample infant; Calculating the acoustic similarity between the target acoustic feature vector and the first acoustic feature vector; When the acoustic similarity is greater than a preset threshold, taking the cry type of the first acoustic feature vector corresponding to the first training sample infant as the cry type of the target infant; When the similarity is less than or equal to the preset threshold, determining a second acoustic feature vector with the highest acoustic similarity to the target acoustic feature vector based on the target acoustic feature vector by using acoustic feature clustering; Taking the cry type corresponding to the second acoustic feature vector as the cry type of the target infant.

2. The method according to claim 1, characterized in that, The establishing the correspondence between the acoustic feature vector, the physiological characteristic parameter, and the cry type specifically includes: Establishing a feature space based on the acoustic feature vector; Mapping the physiological characteristic parameters to the feature space to obtain a physiological feature vector; Calculating the correlation coefficient between the acoustic feature vector and the physiological feature vector; Weighting the acoustic feature vector based on the correlation coefficient to obtain a weighted acoustic feature vector; Establishing the correspondence between the weighted acoustic feature vector and the cry type.

3. The method according to claim 2, wherein The calculating the correlation coefficient between the acoustic feature vector and the physiological feature vector specifically includes: Dividing the acoustic feature vector into multiple feature subsets, and each feature subset corresponds to a feature dimension; Calculating the variance contribution rate of the acoustic feature vector under each feature dimension; Selecting the feature dimensions with the cumulative variance contribution rate greater than a preset contribution rate threshold as the main feature dimensions; Calculating the grey correlation degree between the acoustic feature vector and the physiological feature vector under the main feature dimensions; Calculating the weight coefficient of each main feature dimension based on the grey correlation degree, and taking the weight coefficient as the correlation coefficient.

4. The method according to claim 1, characterized in that After taking the cry type corresponding to the second acoustic feature vector as the cry type of the target infant, the method further includes: Obtaining the satisfaction time points of various physiological needs of the target infant before the current moment; Calculating the time intervals between the current moment and the last satisfaction time points of each physiological need; Establishing a quantization index of the physiological need urgency based on the time intervals; Establishing a probability association between the quantization index and the cry type; When there are multiple candidate cry types, prioritize the candidate cry types based on the probability association to obtain a sorting result; Output the candidate cry type with the highest priority as the final recognition result based on the sorting result.

5. The method according to claim 4, wherein The establishment of a quantitative index of the physiological demand urgency based on the time interval specifically includes: Statistically calculate the average interval period of each physiological demand of the target infant; Establish a non-linear mapping relationship between the time interval and the physiological demand urgency; Calculate the current urgency score of each physiological demand according to the non-linear mapping relationship, and use the current urgency score as the quantitative index.

6. The method according to claim 4, wherein After outputting the candidate cry type with the highest priority as the final recognition result based on the sorting result, the method further includes: Record each recognition result; When the recognition result is incorrect, determine the type of reason that causes the recognition result to be incorrect; Modify the non-linear mapping relationship based on the type of reason.

7. The method according to claim 6, wherein The modification of the non-linear mapping relationship based on the type of reason specifically includes: If the type of reason is an abnormal time interval, adjust the calculation weight of the average interval period; If the type of reason is an incorrect calculation of the demand urgency, optimize the function form of the non-linear mapping relationship.

8. An infant crying monitoring system, characterized in that, The system includes: One or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the system to execute the method according to any one of claims 1-7.

9. A computer-readable storage medium, comprising instructions, characterized in that, When the instructions run on the system, enable the system to execute the method according to any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product runs on the system, enable the system to execute the method according to any one of claims 1-7.

Images