Infant cry appeal recognition system and method

Abstract

The invention relates to the technical field of audio analysis, and discloses an infant cry appeal recognition system and method, and the system comprises a collection module which is configured to collect an infant cry audio in real time, and carries out the preprocessing of the infant cry audio; the feature extraction module is configured to extract the audio duration and the audio spectrum of the infant cry audio; the data processing module is configured to compare the spectrum kurtosis and the spectrum centroid with the historical spectrum kurtosis and the historical spectrum centroid to obtain a plurality of first historical audio spectrums, and obtain a plurality of second historical audio spectrums according to the attenuation degree deviation value; the data processing module is further configured to obtain a plurality of third historical audio frequency spectrums according to the duration deviation value; and the appeal judgment module is configured to identify the crying appeal of the infant according to the second historical audio frequency spectrum and the third historical audio frequency spectrum. According to the method, comprehensive analysis and utilization of multi-dimensional features are realized, and the robustness and the reliability of an identification result are high.

Description

Technical Field

[0001] The present invention relates to the technical field of audio analysis, and in particular to a system and method for recognizing infant crying calls. Background Art

[0002] Infants and young children cry as one of their primary ways of expressing their needs. However, because infants and young children cannot clearly express their needs through language, parents and caregivers often struggle to accurately understand the specific demands behind their cries. Common cries from infants and young children include hunger, sleepiness, discomfort, and pain. Failure to identify and respond promptly can impact the health and emotional development of infants and young children.

[0003] Traditional methods for recognizing infant cries rely primarily on manual experience and simple audio analysis techniques. For example, parents or caregivers assess an infant's needs by listening to the pitch, rhythm, and duration of the cry. However, this approach is subject to significant subjectivity and limitations, and is easily influenced by personal experience and environmental noise, resulting in inaccurate recognition results. Furthermore, with changing family structures and an accelerated pace of life, parents and caregivers have limited time and energy, making it difficult to consistently monitor and respond to infant cries. However, the audio characteristics of infant cries are complex and variable, easily affected by ambient noise and recording equipment, leading to inaccurate feature extraction. Furthermore, the cry characteristics of different infants vary, making a single recognition model unsuitable for all situations. Furthermore, most existing systems rely on a single audio feature and lack the comprehensive analysis and utilization of multidimensional features, resulting in insufficient robustness and reliability in recognition results.

[0004] Therefore, it is necessary to provide a system and method for recognizing infant crying demands to solve the problems that existing crying demand recognition mostly relies on a single audio feature, lacks comprehensive analysis and utilization of multi-dimensional features, and the recognition results are not robust and reliable enough. Summary of the Invention

[0005] In view of this, the present invention proposes a system and method for identifying infant crying demands, aiming to solve the problem that existing crying demand recognition mostly relies on a single audio feature, lacks comprehensive analysis and utilization of multi-dimensional features, and the recognition results are insufficiently robust and reliable.

[0006] In one aspect, the present invention provides a system for identifying infant crying calls, comprising:

[0007] a collection module configured to collect infant crying audio in real time and pre-process the infant crying audio;

[0008] a feature extraction module configured to extract the audio duration and audio spectrum of the infant crying audio, and calculate the spectrum kurtosis, spectrum centroid and spectrum attenuation based on the audio spectrum;

[0009] a data processing module configured to compare the spectrum kurtosis and spectrum centroid with historical spectrum kurtosis and historical spectrum centroid to obtain a plurality of first historical audio spectra, respectively calculate attenuation deviation values between the spectrum attenuation and the historical spectrum attenuation degrees of the plurality of first historical audio spectra, and obtain a plurality of second historical audio spectra based on the attenuation deviation values;

[0010] The data processing module is further configured to calculate duration deviations between the audio duration and the historical audio durations of the plurality of first historical audio spectra, and obtain a plurality of third historical audio spectra according to the duration deviations;

[0011] The appeal judgment module is configured to identify the crying appeal of the infant based on the second historical audio spectrum and the third historical audio spectrum.

[0012] Furthermore, the acquisition module is configured to pre-process the infant crying audio, including:

[0013] Using a filter to remove background noise from the infant crying audio;

[0014] Normalizing the infant crying audio after noise removal to make the volume of the infant crying audio uniform;

[0015] The starting point and the ending point of the infant crying audio are located, and the silent segments are removed to obtain the infant crying audio to be processed.

[0016] Furthermore, when the feature extraction module is configured to extract the audio duration and audio spectrum of the infant crying audio, it includes:

[0017] The spectral kurtosis is calculated by the following formula:

[0018]

[0019] The spectrum centroid is calculated by the following formula:

[0020]

[0021] The spectrum attenuation is calculated by the following formula:

[0022]

[0023] In the above formula, K represents the total number of frequency points in the spectrum; k represents the kth frequency point, where k = 1, 2, 3, ..., K; X(k) represents the spectrum value of the kth frequency point, |X(k)| represents the spectrum amplitude of the kth frequency point, represents the mean of the spectrum amplitude; |X(1)| represents the spectrum amplitude of the first frequency point.

[0024] Furthermore, the data processing module is configured to compare the spectrum kurtosis and spectrum centroid with historical spectrum kurtosis and historical spectrum centroid to obtain a plurality of first historical audio spectra, including:

[0025] The Euclidean distance between the spectral kurtosis and the spectral centroid and the historical spectral kurtosis and the historical spectral centroid is calculated by the following formula:

[0026]

[0027] In the above formula, D represents the Euclidean distance, Kc represents the spectrum kurtosis, Kh represents the historical spectrum kurtosis, SCc represents the spectrum centroid, and SCh represents the historical spectrum centroid;

[0028] A maximum Euclidean distance value is set, and a historical audio spectrum that is less than or equal to the maximum Euclidean distance value is recorded as a first historical audio spectrum.

[0029] Furthermore, when the data processing module is configured to respectively calculate the attenuation deviation values between the spectrum attenuation and the historical spectrum attenuation of a plurality of first historical audio spectra, the calculation includes:

[0030] respectively calculating attenuation deviations between the spectrum attenuation and historical spectrum attenuations of a plurality of first historical audio spectra, and arranging the attenuation deviations in descending order;

[0031] The first historical audio spectra corresponding to the attenuation deviation values are sequentially sorted according to the descending order of the attenuation deviation values.

[0032] Furthermore, when the data processing module is configured to obtain a plurality of second historical audio spectra according to the attenuation deviation value, the method includes:

[0033] Set the maximum value of attenuation deviation;

[0034] Recording a first historical audio spectrum whose attenuation deviation value is smaller than the maximum attenuation deviation value as a second historical audio spectrum;

[0035] If there is only one second historical audio spectrum, the crying demand is directly classified into the historical crying demand category corresponding to the second historical audio spectrum.

[0036] Furthermore, the data processing module is further configured to respectively calculate the duration deviation values of the audio duration and the historical audio durations of the plurality of first historical audio spectra, including:

[0037] Calculate duration deviations between the audio duration and the historical audio durations of the plurality of first historical audio spectra, and arrange the duration deviations in descending order;

[0038] The first historical audio spectra corresponding to the duration deviation values are sequentially sorted according to the descending order of the duration deviation values.

[0039] Furthermore, the data processing module is further configured to obtain a plurality of third historical audio spectra according to the duration deviation value, including:

[0040] Set the maximum value of duration deviation;

[0041] Recording the first historical audio spectrum whose duration deviation value is smaller than the maximum duration deviation value as the third historical audio spectrum;

[0042] If there is only one third historical audio spectrum, the crying appeal is directly classified into the historical crying appeal category corresponding to the third historical audio spectrum.

[0043] Furthermore, the demand judgment module is configured to identify the crying demand of the infant based on the second historical audio spectrum and the third historical audio spectrum, including:

[0044] If there is only one crying demand category corresponding to the second historical audio spectrum and the third historical audio spectrum, the crying demand is classified into the crying demand category corresponding to the second historical audio spectrum and the third historical audio spectrum;

[0045] If there is more than one corresponding crying demand category in the second historical audio spectrum and the third historical audio spectrum, the crying demand is classified into the crying demand category corresponding to the minimum attenuation deviation value in the second historical audio spectrum.

[0046] Compared with the prior art, the beneficial effects of the present invention are that the present invention includes an acquisition module, a feature extraction module, a data processing module and a demand judgment module, which is comprehensive and systematic. First, the acquisition module can collect the crying audio of infants and young children in real time and perform preprocessing to ensure the quality and consistency of the data. Secondly, the feature extraction module provides multi-dimensional feature data for subsequent analysis by extracting the audio duration and audio spectrum, and calculating the spectrum kurtosis, spectrum centroid and spectrum attenuation. The data processing module compares historical data, screens out similar historical audio spectra, and calculates the attenuation deviation value and duration deviation value to further narrow the recognition range. Finally, the demand judgment module identifies the crying demands of infants and young children based on the screening results. This systematic design not only improves the accuracy of recognition, but also enhances the practicality and scalability of the system.

[0047] On the other hand, the present application also provides a method for identifying infant crying demands, comprising:

[0048] Collecting infant crying audio in real time and preprocessing the infant crying audio;

[0049] Extracting the audio duration and audio spectrum of the infant crying audio, and calculating the spectrum kurtosis, spectrum centroid, and spectrum attenuation based on the audio spectrum;

[0050] Comparing the spectrum kurtosis and spectrum centroid with historical spectrum kurtosis and historical spectrum centroid to obtain a plurality of first historical audio spectra, calculating attenuation deviation values between the spectrum attenuation and the historical spectrum attenuation degrees of the plurality of first historical audio spectra, and obtaining a plurality of second historical audio spectra based on the attenuation deviation values;

[0051] Calculating duration deviations between the audio duration and the historical audio durations of the plurality of first historical audio spectra, and obtaining a plurality of third historical audio spectra according to the duration deviations;

[0052] The crying demands of the infant are identified according to the second historical audio spectrum and the third historical audio spectrum.

[0053] It is understandable that the infant crying demand recognition system and method provided in this application have the same beneficial effects and will not be described in detail here. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0055] Figure 1This is a functional block diagram of the infant cry appeal recognition system provided by an embodiment of the present invention;

[0056] Figure 2 This is a flow chart of a method for identifying infant crying demands provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, unless there is a conflict, the embodiments of the present disclosure and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0058] In some embodiments of this application, see Figure 1 As shown, this embodiment provides a system for recognizing infant crying sounds, including:

[0059] The acquisition module is configured to collect the infant crying audio in real time and pre-process the infant crying audio;

[0060] a feature extraction module configured to extract the audio duration and audio spectrum of the infant crying audio, and calculate the spectrum kurtosis, spectrum centroid, and spectrum attenuation based on the audio spectrum;

[0061] a data processing module configured to compare the spectrum kurtosis and the spectrum centroid with the historical spectrum kurtosis and the historical spectrum centroid to obtain a plurality of first historical audio spectra, calculate attenuation deviation values of the spectrum attenuation and the historical spectrum attenuation of the plurality of first historical audio spectra, and obtain a plurality of second historical audio spectra based on the attenuation deviation values;

[0062] The data processing module is further configured to respectively calculate duration deviations between the audio duration and the historical audio durations of the plurality of first historical audio spectra, and obtain a plurality of third historical audio spectra according to the duration deviations;

[0063] The appeal judgment module is configured to identify the crying appeal of the infant based on the second historical audio spectrum and the third historical audio spectrum.

[0064] It can be understood that the present invention includes an acquisition module, a feature extraction module, a data processing module and a demand judgment module, which is comprehensive and systematic. First, the acquisition module can collect infant crying audio in real time and perform preprocessing to ensure the quality and consistency of the data. Secondly, the feature extraction module provides multi-dimensional feature data for subsequent analysis by extracting the audio duration and audio spectrum, and calculating the spectrum kurtosis, spectrum centroid and spectrum attenuation. The data processing module compares historical data, screens out similar historical audio spectra, and calculates the attenuation deviation value and duration deviation value to further narrow the recognition range. Finally, the demand judgment module identifies the infant crying demand based on the screening results. This systematic design not only improves the accuracy of recognition, but also enhances the practicality and scalability of the system.

[0065] Specifically, the crying complaint classification includes hunger, sleepiness, discomfort, and pain.

[0066] In some embodiments of the present application, the acquisition module is configured to pre-process the infant crying audio, including:

[0067] Use filters to remove background noise from audio of babies crying;

[0068] Normalize the infant crying audio after removing the noise to make the volume of the infant crying audio uniform;

[0069] The starting point and ending point of the infant crying audio are located, and the silent segments are removed to obtain the infant crying audio to be processed.

[0070] As you can understand, the acquisition module preprocesses the infant crying audio, including denoising, normalizing, and removing silence segments. Denoising eliminates background noise through filters to ensure audio signal clarity; normalization unifies the volume to avoid recognition errors caused by volume differences; and removing silence segments uses endpoint detection technology to precisely locate the active crying area, reducing the processing of invalid data. These preprocessing steps provide high-quality input for subsequent feature extraction and data processing, significantly improving system stability and recognition accuracy while reducing the impact of environmental noise and device variations on system performance.

[0071] In some embodiments of the present application, when the feature extraction module is configured to extract the audio duration and audio spectrum of the infant crying audio, it includes:

[0072] The spectral kurtosis is calculated by the following formula:

[0073]

[0074] The spectrum centroid is calculated by the following formula:

[0075]

[0076] The spectrum attenuation is calculated by the following formula:

[0077]

[0078] In the above formula, K represents the total number of frequency points in the spectrum; k represents the kth frequency point, where k = 1, 2, 3, ..., K; X(k) represents the spectrum value of the kth frequency point, |X(k)| represents the spectrum amplitude of the kth frequency point, represents the mean of the spectrum amplitude; |X(1)| represents the spectrum amplitude of the first frequency point.

[0079] It is understandable that the feature extraction module comprehensively describes the spectral characteristics of crying by calculating the spectral kurtosis, spectral centroid, and spectral attenuation. Spectral kurtosis reflects the sharpness of the spectral energy distribution, the spectral centroid describes the area of energy concentration, and the spectral attenuation reflects the energy attenuation trend with frequency. These features can effectively distinguish different types of crying and provide multi-dimensional data support for appeal identification. Through specific calculation formulas and parameter definitions, the feature extraction module not only improves the accuracy of feature extraction, but also enhances the repeatability and verifiability of the system, laying a solid foundation for subsequent data processing and identification.

[0080] In some embodiments of the present application, the data processing module is configured to compare the spectrum kurtosis and the spectrum centroid with the historical spectrum kurtosis and the historical spectrum centroid to obtain a plurality of first historical audio spectra, including:

[0081] The Euclidean distance between the spectral kurtosis and spectral centroid and the historical spectral kurtosis and historical spectral centroid is calculated by the following formula:

[0082]

[0083] In the above formula, D represents the Euclidean distance, Kc represents the spectrum kurtosis, Kh represents the historical spectrum kurtosis, SCc represents the spectrum centroid, and SCh represents the historical spectrum centroid;

[0084] The maximum value of the Euclidean distance is set, and a historical audio spectrum that is less than or equal to the maximum value of the Euclidean distance is recorded as the first historical audio spectrum.

[0085] It is understandable that the data processing module screens out similar first historical audio spectra by calculating the spectral kurtosis and spectral centroid of the current audio and the Euclidean distance with the historical data. Euclidean distance can quantify the differences between features and provide an objective basis for similarity comparison. By setting the maximum value of the Euclidean distance, the system can effectively control the screening accuracy and avoid too much or too little similar data. This Euclidean distance-based comparison method not only improves data processing efficiency, but also enhances the system's recognition accuracy, providing a reliable foundation for the subsequent calculation of attenuation deviation and duration deviation values.

[0086] In some embodiments of the present application, when the data processing module is configured to respectively calculate the attenuation deviation values of the spectrum attenuation and the historical spectrum attenuation of a plurality of first historical audio spectra, the calculation includes:

[0087] Calculating attenuation deviations between the spectrum attenuation and the historical spectrum attenuation of the first historical audio spectra, and arranging the attenuation deviations in descending order;

[0088] The first historical audio spectra corresponding to the attenuation deviation values are sequentially sorted according to the descending order of the attenuation deviation values.

[0089] In some embodiments of the present application, when the data processing module is configured to obtain a plurality of second historical audio spectra according to the attenuation deviation value, the process includes:

[0090] Set the maximum value of attenuation deviation;

[0091] Recording the first historical audio spectrum whose attenuation deviation value is smaller than the maximum attenuation deviation value as the second historical audio spectrum;

[0092] If there is only one second historical audio spectrum, the crying demand is directly classified into the historical crying demand category corresponding to the second historical audio spectrum.

[0093] As can be understood, the data processing module calculates the deviation between the spectral attenuation of the current audio and the attenuation of the historical data and sorts them in descending order. By quantifying the difference in spectral attenuation, the system can more accurately screen similar historical audio spectra. The descending sort results intuitively demonstrate the degree of similarity between the current audio and the historical audio, facilitating subsequent screening and processing. This detailed calculation procedure and sorting method not only improves the transparency of data processing but also enhances the system's operability and maintainability, providing an important basis for final voice recognition. Based on the attenuation deviation value, the data processing module selects the second historical audio spectrum. By setting a maximum attenuation deviation value, the system can effectively control the screening accuracy and ensure that the screening results have a high degree of similarity. If there is only one second historical audio spectrum, the system directly classifies the crying voice into the corresponding historical category, simplifying the processing process. This clear screening criteria and processing logic not only improves the system's recognition accuracy but also enhances its practicality and user-friendliness, providing parents and caregivers with an efficient and convenient auxiliary tool.

[0094] In some embodiments of the present application, the data processing module is further configured to calculate the duration deviation values of the audio duration and the historical audio durations of the plurality of first historical audio spectra, respectively, including:

[0095] Calculate duration deviations between the audio duration and the historical audio durations of the plurality of first historical audio spectra, and arrange the duration deviations in descending order;

[0096] The first historical audio spectra corresponding to the duration deviation values are sequentially sorted according to the descending order of the duration deviation values.

[0097] In some embodiments of the present application, the data processing module is further configured to obtain a plurality of third historical audio spectra according to the duration deviation value, including:

[0098] Set the maximum value of duration deviation;

[0099] Recording the first historical audio spectrum whose duration deviation value is smaller than the maximum duration deviation value as the third historical audio spectrum;

[0100] If there is only one third historical audio spectrum, the crying demand is directly classified into the historical crying demand category corresponding to the third historical audio spectrum.

[0101] It is understandable that the data processing module calculates the deviation between the duration of the current audio and the duration of the historical audio, and arranges them in descending order. By quantifying the difference in duration, the system can more accurately screen out similar historical audio spectra. The results of the descending order intuitively show the degree of similarity between the current audio and the historical audio, facilitating subsequent screening and processing. This detailed calculation step and sorting method not only improves the transparency of data processing, but also enhances the operability and maintainability of the system, providing an important basis for the final demand identification. Based on the duration deviation value, the data processing module screens out the third historical audio spectrum. By setting the maximum duration deviation, the system can effectively control the screening accuracy and ensure that the screening results have a high degree of similarity. If there is only one third historical audio spectrum, the system directly divides the crying demand into the corresponding historical category, simplifying the processing process. This clear screening standard and processing logic not only improves the system's recognition accuracy, but also enhances the system's practicality and user-friendliness, providing parents and caregivers with an efficient and convenient auxiliary tool.

[0102] In some embodiments of the present application, the demand determination module is configured to identify the crying demand of an infant or young child based on the second historical audio spectrum and the third historical audio spectrum, including:

[0103] If there is only one crying demand category corresponding to the second historical audio spectrum and the third historical audio spectrum, the crying demand is classified into the crying demand category corresponding to the second historical audio spectrum and the third historical audio spectrum;

[0104] If there is more than one corresponding crying demand category in the second historical audio spectrum and the third historical audio spectrum, the crying demand is classified into the crying demand category corresponding to the minimum attenuation deviation value in the second historical audio spectrum.

[0105] It can be understood that the appeal judgment module identifies the infant's crying appeal based on the results of the second and third historical audio spectra. If only one category corresponds to the same category, the system directly classifies the crying appeal into that category. If multiple categories correspond, the category with the smallest attenuation deviation value is selected. This not only improves the system's recognition accuracy, but also enhances its practicality and user-friendliness, providing parents and caregivers with an efficient and convenient auxiliary tool.

[0106] On the other hand, see Figure 2 As shown, the present application also provides a method for identifying infant crying demands, which is applied to the above-mentioned infant crying demands identification system, comprising the following steps:

[0107] S100, collecting infant crying audio in real time and pre-processing the infant crying audio;

[0108] S200, extracting the audio duration and audio spectrum of the infant crying audio, and calculating the spectrum kurtosis, spectrum centroid, and spectrum attenuation based on the audio spectrum;

[0109] S300: Compare the spectrum kurtosis and the spectrum centroid with the historical spectrum kurtosis and the historical spectrum centroid to obtain a plurality of first historical audio spectra, calculate the attenuation deviation values of the spectrum attenuation and the historical spectrum attenuation of the plurality of first historical audio spectra, and obtain a plurality of second historical audio spectra based on the attenuation deviation values;

[0110] S400: Calculate duration deviations between the audio duration and the historical audio durations of the plurality of first historical audio spectra, and obtain a plurality of third historical audio spectra based on the duration deviations.

[0111] S500: Identify the crying demands of the infant based on the second historical audio spectrum and the third historical audio spectrum.

[0112] It is understandable that this method achieves a comprehensive analysis of infants' crying demands through real-time data collection, preprocessing, feature extraction, data processing, and demand identification. Its advantages lie in its systematicity and practicality, covering the entire process from data collection to demand identification. Through multi-dimensional feature extraction and similarity comparison, the method can more accurately capture the characteristics of crying and significantly improve recognition accuracy. At the same time, the method is also highly scalable and can adapt to the individual differences of different infants and young children, providing parents and caregivers with intelligent auxiliary tools, reducing the manual burden and improving response efficiency.

[0113] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or a combination of software and hardware embodiments. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0114] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0115] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0116] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A system for identifying infant crying demands, characterized in that: include: a collection module configured to collect infant crying audio in real time and pre-process the infant crying audio; a feature extraction module configured to extract the audio duration and audio spectrum of the infant crying audio, and calculate the spectrum kurtosis, spectrum centroid and spectrum attenuation based on the audio spectrum; a data processing module configured to compare the spectrum kurtosis and spectrum centroid with historical spectrum kurtosis and historical spectrum centroid to obtain a plurality of first historical audio spectra, respectively calculate attenuation deviation values between the spectrum attenuation and the historical spectrum attenuation degrees of the plurality of first historical audio spectra, and obtain a plurality of second historical audio spectra based on the attenuation deviation values; The data processing module is further configured to calculate duration deviations between the audio duration and the historical audio durations of the plurality of first historical audio spectra, and obtain a plurality of third historical audio spectra according to the duration deviations; The appeal judgment module is configured to identify the crying appeal of the infant based on the second historical audio spectrum and the third historical audio spectrum.

2. The infant cry recognition system according to claim 1, characterized in that: The acquisition module, when configured to pre-process the infant crying audio, includes: Using a filter to remove background noise from the infant crying audio; Normalizing the infant crying audio after noise removal to make the volume of the infant crying audio uniform; The starting point and the ending point of the infant crying audio are located, and the silent segments are removed to obtain the infant crying audio to be processed.

3. The infant cry recognition system according to claim 2, characterized in that: When the feature extraction module is configured to extract the audio duration and audio spectrum of the infant crying audio, it includes: The spectral kurtosis is calculated by the following formula: The spectrum centroid is calculated by the following formula: The spectrum attenuation is calculated by the following formula: In the above formula, K represents the total number of frequency points in the spectrum; k represents the kth frequency point, where k = 1, 2, 3, ..., K; X(k) represents the spectrum value of the kth frequency point, |X(k)| represents the spectrum amplitude of the kth frequency point, represents the mean of the spectrum amplitude; |X(1)| represents the spectrum amplitude of the first frequency point.

4. The infant cry recognition system according to claim 3, characterized in that: The data processing module is configured to compare the spectrum kurtosis and the spectrum centroid with the historical spectrum kurtosis and the historical spectrum centroid to obtain a plurality of first historical audio spectra, including: The Euclidean distance between the spectral kurtosis and the spectral centroid and the historical spectral kurtosis and the historical spectral centroid is calculated by the following formula: In the above formula, D represents the Euclidean distance, Kc represents the spectrum kurtosis, Kh represents the historical spectrum kurtosis, SCc represents the spectrum centroid, and SCh represents the historical spectrum centroid; A maximum Euclidean distance value is set, and a historical audio spectrum that is less than or equal to the maximum Euclidean distance value is recorded as a first historical audio spectrum.

5. The infant crying demand recognition system according to claim 4, characterized in that: When the data processing module is configured to respectively calculate the attenuation deviation values of the spectrum attenuation and the historical spectrum attenuation of a plurality of first historical audio spectra, the method includes: respectively calculating attenuation deviations between the spectrum attenuation and historical spectrum attenuations of a plurality of first historical audio spectra, and arranging the attenuation deviations in descending order; The first historical audio spectra corresponding to the attenuation deviation values are sequentially sorted according to the descending order of the attenuation deviation values.

6. The infant cry recognition system according to claim 5, characterized in that: When the data processing module is configured to obtain a plurality of second historical audio spectra according to the attenuation deviation value, the method includes: Set the maximum value of attenuation deviation; Recording a first historical audio spectrum whose attenuation deviation value is smaller than the maximum attenuation deviation value as a second historical audio spectrum; If there is only one second historical audio spectrum, the crying demand is directly classified into the historical crying demand category corresponding to the second historical audio spectrum.

7. The infant cry recognition system according to claim 6, characterized in that: The data processing module is further configured to respectively calculate the duration deviation values of the audio duration and the historical audio durations of a plurality of first historical audio spectra, including: Calculate duration deviations between the audio duration and the historical audio durations of the plurality of first historical audio spectra, and arrange the duration deviations in descending order; The first historical audio spectra corresponding to the duration deviation values are sequentially sorted according to the descending order of the duration deviation values.

8. The infant cry recognition system according to claim 7, characterized in that: The data processing module is further configured to obtain a plurality of third historical audio spectra according to the duration deviation value, including: Set the maximum value of duration deviation; Recording the first historical audio spectrum whose duration deviation value is smaller than the maximum duration deviation value as the third historical audio spectrum; If there is only one third historical audio spectrum, the crying appeal is directly classified into the historical crying appeal category corresponding to the third historical audio spectrum.

9. The infant cry recognition system according to claim 8, characterized in that: The demand determination module is configured to identify the crying demand of the infant based on the second historical audio spectrum and the third historical audio spectrum, including: If there is only one crying demand category corresponding to the second historical audio spectrum and the third historical audio spectrum, the crying demand is classified into the crying demand category corresponding to the second historical audio spectrum and the third historical audio spectrum; If there is more than one corresponding crying demand category in the second historical audio spectrum and the third historical audio spectrum, the crying demand is classified into the crying demand category corresponding to the minimum attenuation deviation value in the second historical audio spectrum.

10. A method for recognizing infant crying demands, applied to an infant crying demands recognition system according to any one of claims 1 to 9, characterized in that: include: Collecting infant crying audio in real time and preprocessing the infant crying audio; Extracting the audio duration and audio spectrum of the infant crying audio, and calculating the spectrum kurtosis, spectrum centroid, and spectrum attenuation based on the audio spectrum; Comparing the spectrum kurtosis and spectrum centroid with historical spectrum kurtosis and historical spectrum centroid to obtain a plurality of first historical audio spectra, calculating attenuation deviation values between the spectrum attenuation and the historical spectrum attenuation degrees of the plurality of first historical audio spectra, and obtaining a plurality of second historical audio spectra based on the attenuation deviation values; Calculating duration deviations between the audio duration and the historical audio durations of the plurality of first historical audio spectra, and obtaining a plurality of third historical audio spectra according to the duration deviations; The crying demands of the infant are identified according to the second historical audio spectrum and the third historical audio spectrum.

Images