Infant cry feature extraction method

Abstract

The invention provides a baby cry feature extraction method. The method comprises the following steps: acquiring an audio signal of a baby; preprocessing the audio signal, wherein the preprocessing comprises denoising and volume standardization; analyzing a crying sound waveform in the audio signal, determining a crying sound period, and identifying an expiration stage S1, an inspiration stage S2, an excitation stage and a resting stage; and calculating and extracting the duration, intensity and frequency characteristic parameters of each cry period.

Description

Technical Field

[0001] The present invention relates to the technical field of infant crying sound extraction, and in particular to a method for extracting infant crying sound features. Background Art

[0002] In the field of infant care and monitoring, infant crying, as the most direct and frequent way to express emotions and needs, has always been the focus of researchers and medical staff. Traditional methods of infant crying recognition mostly rely on manual monitoring and judgment, which is not only time-consuming and labor-intensive, but also easily affected by subjective factors, environmental noise, and individual differences of infants, resulting in low recognition accuracy and poor real-time performance.

[0003] With the rapid development of digital signal processing technology and artificial intelligence technology, especially the widespread application of deep learning algorithms, new ideas and methods have been provided for the extraction and recognition of baby crying features. Deep learning can automatically learn and extract high-level abstract features from a large amount of data by building a multi-level neural network model. These features have important guiding significance for subsequent classification, recognition and other tasks.

[0004] In terms of baby cry feature extraction, existing technologies have made some progress, but there are still some challenges. First, baby cries are diverse, and the cry features of different babies in different situations may vary greatly, which makes it difficult to build a universal cry feature extraction model. Secondly, environmental noise is one of the important factors affecting the accuracy of cry feature extraction, especially in complex environments such as homes and hospitals, where noise interference is particularly serious. In addition, individual differences among babies are also factors that cannot be ignored, including age, gender, health status, etc., which will affect the characteristic performance of crying to a certain extent. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a method for extracting baby cry features, which solves the problems of low accuracy and poor real-time performance in the existing baby cry feature extraction methods. By introducing advanced technologies such as deep learning, the accuracy and real-time performance of baby cry recognition are improved, thereby contributing to the development of the field of baby care and monitoring.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A method for extracting baby crying features comprises the following steps: Obtaining audio signals from the baby; Preprocess the audio signal, including denoising and volume normalization; Analyze the crying waveform in the audio signal, determine the crying cycle, and identify the exhalation phase S1, the inhalation phase S2, the excitation phase and the rest phase; The duration, intensity and frequency characteristic parameters of each crying cycle were calculated and extracted.

[0007] Preferably, the pre-processing step comprises applying a fast Fourier transform to perform spectrum analysis on the audio signal to filter out high frequency and low frequency noise.

[0008] Preferably, the step of determining a complete crying cycle comprises: Identify and record the start and end time points of the S1 phase; Identify and record the start and end time points of the S2 phase; An adjacent S1 phase and a subsequent S2 phase are combined into a complete crying cycle; The starting and ending points of the crying cycle were determined by detecting the time interval between consecutive S1 and S2 stages.

[0009] Preferably, identifying the exhalation phase S1 and the inhalation phase S2 comprises: The part with large amplitude changes in the waveform is identified as the exhalation phase S1; The smaller amplitude waveform changes are identified as the S2 phase of the inspiration phase; Excitation and quiescence periods were distinguished based on the periodic characteristics of the waveform.

[0010] Preferably, the method further comprises segmenting the audio signal according to the determined crying cycle, and the segmenting step comprises: Based on the identified S1 and S2 phases, the audio signal is segmented into independent crying cycle segments; Each segmented crying cycle segment is taken as a separate analysis unit for feature extraction; The segmented audio segments are stored and labeled for subsequent analysis and comparison.

[0011] Preferably, the feature extraction step comprises calculating the following parameters: Maximum sound pressure level, minimum sound pressure level, average sound pressure level; The starting value, maximum value, minimum value and average value of the fundamental frequency; The time ratio of the expiratory phase S1 and the inspiratory phase S2; Standard deviation of cry cycles.

[0012] Preferably, the method further comprises the steps of: The extracted crying feature parameters were correlated with the infant's age, gender, health status and external stimulation conditions to establish a prediction model.

[0013] Preferably, the method further comprises exporting the extracted characteristic parameters into a spreadsheet, a database or other data format for subsequent statistical analysis and research.

[0014] A system for executing any one of the methods of claim 1 , comprising: An audio collection module, used to collect audio data of a baby's crying; A data processing module, used to perform preprocessing, feature extraction and classification analysis of audio signals; Output module, used to display or store analysis results.

[0015] Preferably, the computer program product comprises instructions stored in a computer-readable medium, which, when executed by a processor, enables the computer to perform the steps of any one of the methods of claims 1 to 6.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention realizes efficient and accurate extraction of infant crying signals by constructing an optimized deep neural network model and combining advanced signal processing technology and feature extraction algorithm.

[0017] In the data collection and preprocessing stage, the present invention adopts advanced noise reduction technology and silent segment removal algorithm, which effectively reduces the influence of environmental noise and individual differences on crying feature extraction.

[0018] In the model training and optimization stage, the present invention adopts strategies such as cross-validation to ensure the generalization ability of the model on different data sets; finally, in the feature output and application stage, the extracted infant crying features can be used for subsequent crying recognition, thereby analyzing the sound characteristics of infants with different levels of brain development, providing strong support for infant care and monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 The baby crying waveform diagram of the present invention; Figure 2 The waveform diagram of the wow sound exhalation and the hmm sound inhalation of the baby of the present invention; Figure 3 It is a waveform diagram of the stable period of infant crying of the present invention; Figure 4 The waveform diagram of the baby crying excitation period of the present invention; Figure 5 The waveform diagram of the baby audio of the present invention containing multiple S1; Figure 6 A flowchart of the baby audio analysis process of the present invention; Figure 7 The waveform diagram and spectrum diagram of the baby audio of the present invention; Figure 8 Insertion view of the labels of the S1 and S2 regions in the baby audio waveform diagram of the present invention; Fig. 9 Insertion view of labels for the excitation period and the stable period regions in the baby audio waveform diagram of the present invention; Fig.10 A data extraction view of a label track in the baby audio waveform diagram of the present invention; Fig.11 This is the calculation formula for the duration of the baby crying cycle of the present invention. DETAILED DESCRIPTION

[0021] The present invention is further described in detail below with reference to the accompanying drawings.

[0022] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations. The basic principles of the present invention defined in the following description can be used for other embodiments, variations, improvements, equivalents, and other technical solutions that do not deviate from the spirit and scope of the present invention.

[0023] Those skilled in the art should understand that, in the disclosure of the present invention, the directions or positions indicated by the terms "longitudinal", "lateral", "up", "down", "left", "right", "front", "back", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc. are based on the directions or positional relationships shown in the accompanying drawings, which are only simplified descriptions for the convenience of describing the present invention, and do not indicate or imply that the device or component referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, the above terms should not be understood as limitations on the present invention.

[0024] It is to be understood that the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element may be one, while in another embodiment, the number of the element may be multiple, and the term "one" should not be understood as a limitation on the quantity.

[0025] The present invention is further described in detail below in conjunction with the accompanying drawings: See also Figure 1-Figure 11 The present invention provides a method for extracting characteristics of a baby's cry, comprising the following steps: 1) Obtain the baby’s audio signal; 2) Preprocess the audio signal, including denoising and volume normalization; 3) Analyze the crying waveform in the audio signal, determine the crying cycle, and identify the exhalation phase S1, the inhalation phase S2, the excitation phase and the rest phase; 4) Calculate and extract the duration, intensity and frequency characteristic parameters of each crying cycle.

[0026] When collecting the audio signal of the baby in the present invention, the baby's crying sound is collected by a recording device. During recording, the recording device is fixed as close to the baby's mouth or chest as possible to reduce the interference of environmental noise and increase the intensity of the crying signal.

[0027] The data processing module in the present invention preprocesses the audio signal, and the preprocessing step includes applying fast Fourier transform to perform spectrum analysis on the audio signal to filter out high-frequency and low-frequency noise: Applying Fast Fourier Transform (FFT): Converts the signal from the time domain to the frequency domain, making it easier to analyze the frequency content of the signal; Spectrum analysis: Calculate the amplitude of each frequency component from the FFT result. By visualizing the spectrum, you can identify which frequency bands contain noise (high frequency or low frequency). Noise removal: A bandpass filter can be used to filter out high-frequency and low-frequency noise by allowing signals within a certain frequency range to pass through and attenuating frequencies outside this range; Volume Normalization: The process of adjusting the volume of an audio signal to a predetermined, relatively constant level.

[0028] The steps of determining a complete crying cycle in the present invention include: Identify and record the start and end time points of the S1 phase; Identify and record the start and end time points of the S2 phase; See also Figure 1 , combining an adjacent S1 phase and a subsequent S2 phase into a complete crying cycle; By detecting the time interval between consecutive S1 and S2 stages, the start and end of the crying cycle can be determined; Babies often cry when they are just born, and their crying has a certain pattern (related to breathing). When the baby cries and makes a "wa" sound, the baby exhales, and this period is recorded as S1. When the baby makes a "hmm" sound similar to hiccups after crying once, the baby inhales, and this period is recorded as S2. Since the baby cries hard for many times in a row (exhales) after a short rest (inhales), there are multiple S1s, so the total time period of multiple hard crying is recorded as the excitation period (that is, one excitation period can include multiple S1s). Similarly, the baby rests (inhales) many times after crying relatively steadily (exhaling). At this time, the total time period of multiple rests is recorded as the resting period (that is, one resting period can include multiple S2s). Therefore, a crying cycle = excitation period + resting period.

[0029] The present invention includes the following steps to identify the exhalation phase S1 and the inhalation phase S2: The part with large amplitude changes in the waveform is identified as the exhalation phase S1; The smaller amplitude waveform changes are identified as the S2 phase of the inspiration phase; Distinguish between excitation and quiescence periods based on the periodic characteristics of the waveform; The exhalation phase S1 and the inhalation phase S2 are determined based on the waveform changes in the waveform diagram. The waveform diagram with a large amplitude change is the exhalation phase S1, and the waveform diagram with a small amplitude change is the inhalation phase S2.

[0030] See also Figure 2-Figure 9 , further, segmenting the audio signal according to the determined crying cycle, the segmentation step comprising: Based on the identified S1 and S2 phases, the audio signal is segmented into independent crying cycle segments; Each segmented crying cycle segment is taken as a separate analysis unit for feature extraction; Store and label the segmented audio segments for subsequent analysis and comparison; The key point of separation is that when playing the baby audio sound, it can be found that a baby with S1+S2 often makes a wow\hmm sound. The gap between S1 and S2 (there is a relatively stable baseline) can be understood as the baby's breathing. As long as it is disconnected from S1, it is S2. The one between two S1s is S2. S1 often has a larger and longer waveform, while S2 has a smaller and shorter waveform. Determination of a crying cycle: find similar adjacent S1 in the wave graph. Two adjacent crying graphs with high similarity can be judged as S1, d(S1-S2)>d(S2-S1), the frequency of S1 is higher than that of S2, and the highest peak of S1 is higher than the highest peak of S2. Select two S1 time periods and add crying cycle labels to the selected time periods.

[0031] When performing feature extraction, the feature extraction step includes calculating the following parameters: Maximum sound pressure level, minimum sound pressure level, average sound pressure level, where the maximum sound pressure level is the highest intensity of the sound in a given time period; the minimum sound pressure level is the lowest intensity of the sound in a given time period; the average sound pressure level is the average intensity of the sound in a given time period, calculated by the energy averaging method; The starting value, maximum value, minimum value and average value of the fundamental frequency, where the starting value of the fundamental frequency is the first fundamental frequency value detected at the beginning of the analyzed speech signal; the maximum value of the fundamental frequency is the highest fundamental frequency value detected in the analyzed speech signal segment; the minimum value of the fundamental frequency corresponds to the maximum value and is the lowest fundamental frequency value detected in the analyzed speech signal segment; the average value of the fundamental frequency is the arithmetic average of all detected fundamental frequency values ​​in the analyzed speech signal segment; The time ratio of the expiratory phase S1 and the inspiratory phase S2; Standard deviation of the crying cycle: reflects the stability or variability of the crying signal between different cycle segments.

[0032] The present invention is based on a method for extracting characteristics of infant crying sounds, and further comprises the following steps: The extracted cry characteristic parameters were correlated with the baby's age, gender, health status and external stimulation conditions to establish a prediction model; See also Figure 10-11 ,Furthermore, it includes exporting the extracted characteristic parameters into a spreadsheet, a database or other data format for subsequent statistical analysis and research. The characteristic parameter calculation formula is: S1 duration = S1 (end time) - S1 (start time); S2 duration = S2 (end time) - S2 (start time); The duration of the excitation period = the first S2 at the moment (starting time) - the first S1 at the moment (starting time); Resting period duration = next S1 (starting time) - current first S2 (starting time); The duration of the crying cycle = excitation period + quiet period = next S1 (starting time) - current S1 (starting time).

[0033] Based on the above-mentioned baby crying feature extraction method, the following steps are included: Acquisition of infant audio signals through recording equipment; Preprocessing the acquired audio signal; Analyze the crying waveform in the audio signal, determine the crying cycle, and identify the exhalation phase S1, the inhalation phase S2, the excitation phase and the rest phase; Finally, the duration, intensity and frequency characteristic parameters of each crying cycle are calculated and extracted.

[0034] A system based on the above-mentioned method for extracting baby crying features comprises: An audio collection module, used to collect audio data of a baby's crying; A data processing module, used to perform preprocessing, feature extraction and classification analysis of audio signals; Output module, used to display or store analysis results.

[0035] The computer program product includes instructions stored in a computer-readable medium, which, when executed by a processor, enables the computer to perform the steps of any of the above-mentioned methods for extracting baby cry features. Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and the embodiments of the present invention may be deformed or modified in any way without departing from the principles.

Claims

1. A method for extracting baby crying features, comprising the following steps: Get the baby's audio signal; Preprocess the audio signal, including denoising and volume normalization; Analyze the crying waveform in the audio signal, determine the crying cycle, and identify the exhalation phase S1, the inhalation phase S2, the excitation phase and the rest phase; The duration, intensity and frequency characteristic parameters of each crying cycle were calculated and extracted.

2. The method for extracting baby crying features according to claim 1, characterized in that: The pre-processing step includes applying Fast Fourier Transform to perform spectrum analysis on the audio signal to filter out high-frequency and low-frequency noise.

3. The method for extracting baby crying features according to claim 1, characterized in that: The steps to determine a complete crying cycle include: Identify and record the start and end time points of the S1 phase; Identify and record the start and end time points of the S2 phase; An adjacent S1 phase and a subsequent S2 phase are combined into a complete crying cycle; The starting and ending points of the crying cycle were determined by detecting the time interval between consecutive S1 and S2 stages.

4. The method for extracting baby crying features according to claim 3, characterized in that: Identifying the exhalation phase S1 and the inhalation phase S2 phases includes: The part with large amplitude changes in the waveform is identified as the exhalation phase S1; The smaller amplitude waveform changes are identified as the S2 phase of the inspiration phase; Excitation and quiescence periods were distinguished based on the periodic characteristics of the waveform.

5. The method for extracting baby crying features according to claim 4, characterized in that: The method further comprises segmenting the audio signal according to the determined crying cycle, wherein the segmenting step comprises: Based on the identified S1 and S2 phases, the audio signal is segmented into independent crying cycle segments; Each segmented crying cycle segment is taken as a separate analysis unit for feature extraction; The segmented audio segments are stored and labeled for subsequent analysis and comparison.

6. The method for extracting baby crying features according to claim 1, characterized in that: The feature extraction step includes calculating the following parameters: Maximum sound pressure level, minimum sound pressure level, average sound pressure level; The starting value, maximum value, minimum value and average value of the fundamental frequency; The time ratio of the expiratory phase S1 and the inspiratory phase S2; Standard deviation of cry cycles.

7. The method for extracting baby crying features according to claim 6, characterized in that: Also includes the steps: The extracted crying feature parameters were correlated with the infant's age, gender, health status and external stimulation conditions to establish a prediction model.

8. The method for extracting baby crying features according to claim 1, characterized in that: The method further includes exporting the extracted characteristic parameters into a spreadsheet, a database or other data formats for subsequent statistical analysis and research.

9. A system for performing the method of any one of claims 1 to 8, comprising: An audio collection module, used to collect audio data of a baby's crying; A data processing module, used to perform preprocessing, feature extraction and classification analysis of audio signals; Output module, used to display or store analysis results.

10. A computer program product for use in the system of claim 7, characterized in that The computer program product comprises instructions stored in a computer-readable medium, which, when executed by a processor, enable a computer to perform the steps of any one of the methods of claims 1 to 6.

Images