Intelligent audio dynamic control method based on intelligent wearable device

Abstract

The invention discloses an intelligent audio dynamic control method based on intelligent wearable equipment, and relates to the technical field of intelligent wearable technology and audio processing, and the method comprises the following steps: obtaining physiological information and position information of a wearer, and carrying out the preprocessing of the data; determining a data acquisition time interval according to the preprocessed data; according to the heart rate information of the wearer, the state of the wearer is divided into a first type of activity state and a second type of activity state; according to the information of the wearer and the position information, the moving state of the wearer is judged, an information packet P1 is generated in the moving state, and an information packet P2 is generated in the static state; judging whether the wearer is in a moving state or a static state, and when the wearer is in the moving state, carrying out balanced sounding and mega bass adjustment according to the information packet P1; and when the intelligent wearable device is in the static state, setting the sound production weight of the loudspeaker according to the information packet P2, and adjusting the audio by constructing a model.

Description

Technical Field

[0001] The present invention relates to the technical fields of intelligent wearable technology and audio processing technology, and specifically to an intelligent audio dynamic control method based on intelligent wearable devices. Background Art

[0002] With the continuous improvement of people's requirements for the audio experience of intelligent wearable devices, intelligent audio dynamic control plays an increasingly crucial role in enhancing the user experience. As the core technology of the audio function of intelligent wearable devices, the accuracy and adaptability of its control affect the user's audio enjoyment and the satisfaction of device use.

[0003] However, traditional intelligent wearable device audio control methods often face the following problems when dealing with complex usage scenarios, diverse user states, and personalized audio requirements: First, the control lacks real-time performance and accuracy. In different usage scenarios and user states of intelligent wearable devices, the user's audio requirements vary greatly, but traditional methods are difficult to quickly and accurately adjust the audio according to these real-time changing information. Second, the control means are single and difficult to meet diverse requirements. Traditional intelligent wearable device audio control usually only relies on simple preset modes or user manual adjustments, and cannot fully consider the special requirements in complex scenarios. In addition, there is a lack of intelligent learning and adaptive capabilities, and it cannot optimize itself according to the user's historical behavior and feedback. Traditional audio control methods do not make full use of the user's historical usage data, audio adjustment records, and audio effect feedback, cannot learn the user's usage habits and preferences, are difficult to achieve intelligent audio control, and cannot dynamically adjust the audio control strategy in the face of changing usage scenarios and user requirements, and cannot meet the growing intelligent and personalized audio requirements of users. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent audio dynamic control method based on intelligent wearable devices to solve the problems raised in the prior art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: An intelligent audio dynamic control method based on intelligent wearable devices, the method includes the following steps: Obtain the physiological information and location information of the wearer through the intelligent wearable device, and preprocess the data; According to the preprocessed data, draw a historical data trend graph, extract the continuous time intervals corresponding to the maximum and minimum values of the data points, select the minimum value as the selected period, and determine the data acquisition time interval; Based on the heart rate information of the wearer, compare it with the set heart rate threshold, and divide the wearer's state into the first type and the second type of activity states; Judge the movement state of the wearer according to the wearer information and location information, generate information packet P1 for the movement state, and generate information packet P2 for the stationary state; Determine whether the wearer is in a moving or stationary state. When in a moving state, balance the sound and enhance the bass according to the information packet P1. When in a stationary state, set the speaker sound weight according to the information packet P2, and adjust the audio by constructing a model.

[0006] Collect sensor data of the smart wearable device in different usage scenarios, and preprocess the collected sensor data. The specific steps are as follows: In the scenario of wearing the smart wearable device, at the initial sampling time interval, use the sensors built in the smart wearable device to collect the wearer's information. The obtained wearer information includes physiological information and location information, where the physiological information includes heart rate information and blood oxygen information. For heart rate information collection, use a heart rate sensor to collect data and record the heart rate value H within every T seconds. For blood oxygen information collection, use a blood oxygen sensor to collect data and record the blood oxygen saturation value O within every T seconds. In terms of location information collection, use the GPS positioning module built in the smart wearable device to collect location data and record the location coordinate information L within every T seconds. Preprocess the obtained data. After processing by the sliding average filtering method, the preprocessed heart rate data H is obtained ’ , after processing by the median filtering algorithm, the preprocessed blood oxygen saturation data O is obtained ’ , after processing by the Kalman filtering algorithm, the preprocessed location coordinate data L is obtained ’ , perform normalization processing on different types of data, and normalize the heart rate, blood oxygen saturation, and location coordinate data to the interval [0, 1].

[0007] The method for determining the data collection interval is as follows: Obtain the wearer's physiological information and location information collected by the smart wearable device in the past period of time. For each type of data, sort it in chronological order to obtain several historical data points. Mark the historical data points of each type of data in the plane rectangular coordinate system respectively, and connect the adjacent historical data points in chronological order to obtain the historical data trend graph of each data type. For the historical data trend graph of each data type, extract the continuous time intervals corresponding to the minimum and maximum values of the data points. Sort the extracted all continuous time intervals from largest to smallest. For each data type, select the minimum value in the continuous time intervals as the selected period of this data type, and use the selected period to determine the time interval for the smart wearable device to collect this type of data.

[0008] Divide the wearer's state into the first type of activity state and the second type of activity state according to the collected wearer heart rate information; The first type of activity state is defined as follows: for the collected wearer's heart rate greater than the set heart rate threshold, it is determined that the wearer is in the first type of activity state, where the set heart rate threshold is calculated from historical data; The second type of activity state is defined as follows: for the collected wearer's heart rate less than or equal to the set heart rate threshold, it is determined that the wearer is in the second type of activity state.

[0009] Based on the obtained wearer information and location information, the movement behavior of the wearer is judged. First, it is judged whether the wearer is in a moving state. For the wearer in a moving state, the heart rate information of the wearer in the moving state is integrated to generate an information packet P1; for the wearer in a stationary state, the heart rate information and location information of the wearer in the stationary state are integrated together to generate an information packet P2.

[0010] For the wearer in a moving state, first, the built-in speaker of the smart wearable device is evenly sounded according to the information in the information packet P1; Next, it is judged whether the wearer is in the first type of activity state or the second type of activity state. For the wearer in the first type of activity state, the overall output of the speaker matrix is adjusted to a soft hypnotic state by lowering the bass; for the wearer in the second type of activity state, the overall output of the speaker matrix is adjusted to meet the audio style requirements in the second type of activity state by increasing the bass.

[0011] For the wearer in a stationary state, first, the sound weight of the built-in speaker of the smart wearable device is set according to the information in the information packet P2; Next, it is judged whether the wearer is in the first type of activity state or the second type of activity state. For the wearer in the first type of activity state, the overall output of the speaker matrix is adjusted to a soft hypnotic state by lowering the bass and fine-tuning the gain of the frequency bands of each speaker in combination with the acoustic environment where the wearer is located; For the wearer in the second type of activity state, the overall output of the speaker matrix is adjusted to meet the audio style requirements in the second type of activity state by increasing the bass and adjusting the mid-high frequency output of the speakers at different positions in combination with the location information of the wearer.

[0012] The steps for setting the sound weight of the built-in speaker of the smart wearable device according to the information in the information packet P2 are as follows: By collecting acoustic environment information, historical voice weights, audio effect feedback, as well as the physiological information and location information collected above, among which, the acoustic environment information uses the microphone built in the smart wearable device to collect environmental sounds, and analyzes acoustic parameters such as its frequency distribution, volume size, and reverberation time; the historical voice weights and audio effect feedback record the voice weights set for the speaker in the past and the corresponding audio effect evaluations, and the evaluation method is the subjective feedback of the user, and the evaluation methods include scoring and text evaluation, and the indicators obtained through objective audio analysis tools include sound clarity and timbre balance; For heart rate data, extract the statistical features of the mean and standard deviation of the heart rate. For acoustic environment data, extract the energy distribution features of the main frequency components, and select the features with higher correlation with the voice weight as the input of the model through correlation analysis; In terms of model selection, adopt the method of 5-fold cross-validation to evaluate the performance of the linear regression model, decision tree model, and neural network model respectively. For the linear regression model, use the mean squared error MSE as the evaluation index; for the decision tree model, use MSE for evaluation, and at the same time consider the complexity of the model, specifically measured by the depth of the tree. For the neural network model, use the cross-entropy loss function for evaluation. By comparing the performance of different models in cross-validation, select the model with the best performance, and set the voice weight for the speaker built in the smart wearable device according to the voice weight value predicted by the model.

[0013] The specific steps for adjusting the frequency band of the speaker built in the smart wearable device are as follows: Collect the audio data reflecting the states of heart rate, blood oxygen saturation, and location information of the smart wearable device for different wearers, as well as the corresponding user state information and audio adjustment records, and construct a database; Obtain the records in the database. The speaker built in the smart wearable device has n adjustable frequency bands (low frequency band, mid-low frequency band, mid-high frequency band, high frequency band, etc.), a1, a2,..., a n respectively represent the adjustment amounts (gain adjustment values or equalization adjustment parameters) of the 1st, 2nd,..., nth frequency bands in the current record, and t1 represents the change amount of the evaluation index of the adjusted audio effect (the value of improved sound clarity, the value of improved matching degree with the environment, etc.), and obtain the function equation f(a1, a2,..., a n ) = t1; Select a neural network model to generate a multi-variable mapping function, and output the multi-variable mapping function P(x1, x2,..., x n ) between the adjustment amounts of the frequency bands and the change amounts of the audio effect evaluation indicators, where, x1, x2,..., x n respectively represent the adjustment amounts of the 1st, 2nd,..., nth frequency bands; The real-time collected and preprocessed wearer-related data is sorted and transformed according to the input requirements of the multivariate mapping function P(x1, x2,..., x n ) and then input into the function to calculate the predicted frequency band adjustment amounts y1, y2,..., y n , where the predicted frequency band adjustment amounts represent the parameter values that need to be adjusted for each frequency band to achieve the best audio effect.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By dynamically analyzing the changing trends of the wearer's physiological information, location information, and acoustic environment information, and combining the audio requirements in different activity states, the potential audio requirements of the wearer in different scenarios are identified; 2. By combining the smart wearable device with the smart speaker matrix system, by collecting the human heart rate, blood oxygen information, and location information, intelligent audio output control is achieved. Based on data mining, the physical state and movement situation are judged according to the human body data, and the power amplifier gain and output state of the speaker matrix are dynamically adjusted. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic flowchart of a smart audio dynamic control method based on a smart wearable device according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] In the embodiment: As Figure 1 shown, the present invention provides a technical solution, a smart audio dynamic control method based on a smart wearable device, and the method includes the following steps: Obtain the wearer's physiological information and location information through the smart wearable device, and preprocess the data; According to the preprocessed data, draw a historical data trend graph, extract the continuous time intervals corresponding to the maximum and minimum values of the data points, select the minimum value as the selected period, and determine the data collection time interval; According to the wearer's heart rate information, compare it with the set heart rate threshold, and divide the wearer's state into the first type and the second type of activity states; Judge the movement state according to the wearer's information and location information, generate an information packet P1 for the movement state, and generate an information packet P2 for the stationary state; Determine whether the wearer is in a moving or stationary state. When in a moving state, the sound is evenly emitted according to information packet P1 and the bass is emphasized; when in a stationary state, the sound weight of the speaker is set according to information packet P2, and the audio is adjusted by constructing a model.

[0018] Collect sensor data of the smart wearable device in different usage scenarios, and preprocess the collected sensor data. The specific steps are as follows: In the scenario of wearing the smart wearable device, according to the initial sampling time interval, use the sensors built in the smart wearable device to collect data on the wearer's information. The wearer information obtained includes physiological information and location information, where the physiological information includes heart rate information and blood oxygen information. For heart rate information collection, use a heart rate sensor to collect data and record the heart rate value H within every T seconds. For blood oxygen information collection, use a blood oxygen sensor to collect data and record the blood oxygen saturation value O within every T seconds. In terms of location information collection, use the GPS positioning module built in the smart wearable device to collect location data and record the location coordinate information L within every T seconds. Preprocess the obtained data. After processing by the moving average filtering method, the preprocessed heart rate data H is obtained ’ , and after processing by the median filtering algorithm, the preprocessed blood oxygen saturation data O is obtained ’ , and after processing by the Kalman filtering algorithm, the preprocessed location coordinate data L is obtained ’ , and normalize different types of data, and normalize the heart rate, blood oxygen saturation, and location coordinate data to the [0, 1] interval.

[0019] Specifically, collect the wearer's heart rate data at a frequency of 1 time per second through the smart wearable device, collect the blood oxygen saturation data once every 30 seconds, obtain the location coordinate information once per minute through the GPS positioning module, and at the same time, the microphone continuously collects environmental sound data.

[0020] The method for determining the data collection interval is as follows: Obtain the physiological information and location information of the wearer collected by the smart wearable device in the past period of time. For each type of data, sort it in chronological order to obtain a number of historical data points. Mark the historical data points of each type of data in the plane rectangular coordinate system respectively, and connect the adjacent historical data points in chronological order to obtain the historical data trend graph of each data type. For the historical data trend graph of each data type, extract the continuous time intervals corresponding to the minimum and maximum values of the data points. Sort all the extracted continuous time intervals in descending order. For each data type, select the minimum value in the continuous time intervals as the selected period of this data type. Use the selected period to determine the time interval for the smart wearable device to collect this type of data.

[0021] Specifically, obtain the physiological information and location information of the wearer collected by the smart bracelet in the past 24 hours through the smart wearable device. Take the average heart rate per minute as a data point, mark it in the plane rectangular coordinate system after sorting in chronological order, and connect the adjacent data points to obtain the heart rate historical data trend graph. Extract the continuous time intervals corresponding to the minimum and maximum values of the data points from the heart rate data trend graph. According to the time interval corresponding to the minimum value is 3 minutes and the time interval corresponding to the maximum value is 30 minutes, after sorting all the time intervals from large to small, select the minimum value of 3 minutes as the selected period of the heart rate data, that is, adjust to collect the heart rate data every 3 minutes in the future. Similarly, determine the data collection intervals of blood oxygen saturation and location information.

[0022] Divide the wearer's state into the first type of activity state and the second type of activity state according to the collected heart rate information of the wearer; The definition of the first type of activity state is as follows: for the collected heart rate of the wearer greater than the set heart rate threshold, it is judged that the wearer is in the first type of activity state, where the set heart rate threshold is calculated from historical data; The definition of the second type of activity state is as follows: for the collected heart rate of the wearer less than or equal to the set heart rate threshold, it is judged that the wearer is in the second type of activity state.

[0023] Specifically, by analyzing the heart rate data of the past week, calculate that the average heart rate is 75 beats per minute, and set the heart rate threshold to 80 beats per minute. When the collected heart rate of the wearer is greater than 80 beats per minute, it is judged that the wearer is in the first type of activity state; when the heart rate is less than or equal to 80 beats per minute, it is judged that the wearer is in the second type of activity state.

[0024] Based on the obtained wearer information and location information, the movement behavior of the wearer is judged. First, it is judged whether the wearer is in a moving state. When the wearer is in a moving state, the heart rate information of the wearer in the moving state is integrated to generate information packet P1; when the wearer is in a stationary state, the heart rate information and location information of the wearer in the stationary state are integrated together to generate information packet P2.

[0025] Specifically, based on the continuously obtained position coordinate information L ’ The movement state of the wearer is judged. When the position coordinates change by more than the threshold value (latitude and longitude change by more than 0.001 degrees) within 5 minutes, it is judged that the wearer is in a moving state. When the wearer is in a moving state, the heart rate information in the moving state is integrated to generate information packet P1. When the wearer is in a stationary state, the heart rate information and location information in the stationary state are integrated together to generate information packet P2.

[0026] When the wearer is in a moving state, first, the built-in speaker of the smart wearable device is evenly sounded according to the information in information packet P1; Next, it is judged whether the wearer is in the first type of activity state or the second type of activity state. When the wearer is in the first type of activity state, the overall output of the speaker matrix is adjusted to a soft hypnosis state by lowering the heavy bass; when the wearer is in the second type of activity state, the overall output of the speaker matrix is adjusted to meet the audio style requirements in the second type of activity state by increasing the heavy bass.

[0027] Specifically, the built-in speaker of the smart earphone is evenly sounded according to information packet P1. When it is judged that the wearer is in the first type of activity state, by lowering the heavy bass, the gain of the low frequency band (20 - 200 Hz) is reduced by 20%, and the overall output of the speaker matrix is adjusted to a soft hypnosis state. When in the second type of activity state, by increasing the heavy bass, the gain of the low frequency band is increased by 15% to meet the audio style requirements in this state.

[0028] When the wearer is in a stationary state, first, the sound weight of the built-in speaker of the smart wearable device is set according to the information in information packet P2; Next, it is judged whether the wearer is in the first type of activity state or the second type of activity state. When the wearer is in the first type of activity state, by lowering the heavy bass and combining with the acoustic environment where the wearer is located, the gain of the frequency band of each speaker is finely adjusted to adjust the overall output of the speaker matrix to a soft hypnosis state; When the wearer is in the second type of activity state, by increasing the heavy bass and combining with the location information of the wearer, the mid - high frequency output of the speakers at different positions is adjusted to adjust the overall output of the speaker matrix to meet the audio style requirements in the second type of activity state.

[0029] Set the sound - emitting weight of the speaker built into the smart wearable device according to the information in packet P2. The specific steps are as follows: Collect the acoustic environment information, historical sound - emitting weight and audio - effect feedback, as well as the physiological information and location information collected above. Among them, the acoustic environment information uses the microphone built into the smart wearable device to collect environmental sounds, and analyzes its frequency distribution, volume size and reverberation - time acoustic parameters. The historical sound - emitting weight and audio - effect feedback record the sound - emitting weights set for the speaker in the past and the corresponding audio - effect evaluations. The evaluation method is the subjective feedback of the user, including scoring and text evaluation, and the indicators obtained through objective audio - analysis tools, including sound clarity and timbre balance; For heart - rate data, extract the statistical features of the mean and standard deviation of the heart rate. For acoustic - environment data, extract the energy - distribution features of the main frequency components, and select the features with higher correlation with the sound - emitting weight as the input of the model through correlation analysis; In terms of model selection, use the 5 - fold cross - validation method to evaluate the performance of the linear - regression model, decision - tree model and neural - network model respectively. For the linear - regression model, use the mean - squared error (MSE) as the evaluation index; for the decision - tree model, use MSE for evaluation and consider the complexity of the model, specifically measured by the depth of the tree. For the neural - network model, use the cross - entropy loss function for evaluation. By comparing the performance of different models in cross - validation, select the model with the best performance, and set the sound - emitting weight of the speaker built into the smart wearable device according to the sound - emitting weight value predicted by the model.

[0030] Specifically, set the sound - emitting weight of the speaker built into the smart earphone according to the information in packet P2, collect the acoustic environment information, and analyze the frequency distribution, volume size and reverberation - time parameters of the environmental sound through the microphone. The environmental sound has a higher energy in the frequency band of 500 - 1000 Hz, the volume is 60 dB, and the reverberation time is 0.5 seconds; Extract the statistical features of the mean and standard deviation of the heart - rate data. The mean heart rate is 72 beats per minute and the standard deviation is 3 within a certain period of time; Select the features with higher correlation with the sound - emitting weight as the input of the model through correlation analysis, use the 5 - fold cross - validation to evaluate the performance of the linear - regression model, decision - tree model and neural - network model, select the neural - network model, and set the sound - emitting weight of the speaker built into the smart earphone according to the sound - emitting weight value predicted by the model; When the wearer is in the first - type activity state, turn down the bass (the gain in the low - frequency band is reduced by 10%), and fine - tune the gain of each speaker frequency band in combination with the acoustic environment at the location. Since the environmental sound has a higher energy in the frequency band of 500 - 1000 Hz, reduce the gain of the speakers in this frequency band by 5%, and adjust the overall output of the speaker matrix to a soft and hypnotic state.

[0031] When the wearer is in the second type of activity state, the bass is increased (the low-frequency band gain is increased by 12%), and combined with the position information, the mid-high frequency output of the speakers at different positions is adjusted. When it is detected that the wearer is in a small indoor space, the gain of the mid-high frequency band (2000 - 5000 Hz) is reduced by 8%, and the overall output of the speaker matrix is adjusted to meet the requirements for the audio style in this state.

[0032] The specific steps for adjusting the frequency bands of the speakers built into the smart wearable device are as follows: Collect the audio data of the smart wearable device in the states reflected by the heart rate, blood oxygen saturation, and position information of different wearers, as well as the corresponding user state information and audio adjustment records, and construct a database; Obtain the records in the database. There are n adjustable frequency bands (low-frequency band, mid-low frequency band, mid-high frequency band, high-frequency band, etc.) for the speakers built into the smart wearable device, a1, a2,..., a n respectively represent the adjustment amounts (gain adjustment values or equalization adjustment parameters) of the 1st, 2nd,..., nth frequency bands in the current record, and t1 represents the change amount of the evaluation index of the adjusted audio effect (the value of improved sound clarity, the value of improved matching degree with the environment, etc.), and obtain the function equation f(a1, a2,..., a n ) = t1; Select a neural network model to generate a multi-variable mapping function, and output the multi-variable mapping function P(x1, x2,..., x n ) between the adjustment amounts of the frequency bands and the change amount of the evaluation index of the audio effect, where x1, x2,..., x n respectively represent the adjustment amounts of the 1st, 2nd,..., nth frequency bands; Organize and convert the real-time collected and preprocessed data related to the wearer according to the input requirements of the multi-variable mapping function P(x1, x2,..., x n ) and then input it into the function to calculate the predicted adjustment amounts of the frequency bands y1, y2,..., y n , where the predicted adjustment amounts of the frequency bands represent the parameter values that each frequency band needs to be adjusted in order to achieve the best audio effect.

[0033] Specifically, collect the audio data, user state information, and audio adjustment records of the smart wearable device in different wearer states, construct a database, and record that when the heart rate is 80 beats per minute, the blood oxygen saturation is 97%, and the position is in the park, the adjustment amount of the audio low-frequency band is 10, and the value of improved sound clarity after adjustment is 0.5. There are 4 adjustable frequency bands for the speakers built into the smart earphone, which are the low-frequency band, mid-low frequency band, mid-high frequency band, and high-frequency band respectively. a1, a2, a3, a4 respectively represent the adjustment amounts of each frequency band, and t1 represents the change amount of the evaluation index of the adjusted audio effect, and obtain the function equation f(a1, a2,..., an ) = t1, select a neural network model to generate a multi - variable mapping function P(x1, x2,..., x n ) After collecting and pre - processing the wearer - related data in real time, organize and transform it according to the function input requirements and then input it into the function. After inputting the data, calculate the predicted frequency band adjustment amounts y1 = 15 (low - frequency band adjustment amount), y2 = 10 (mid - low - frequency band adjustment amount), y3 = - 5 (mid - high - frequency band adjustment amount), y4 = 3 (high - frequency band adjustment amount). Adjust the frequency band of the intelligent headphone speaker according to these adjustment amounts to achieve the best audio effect.

[0034] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above - mentioned exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non - restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

Claims

1. An intelligent audio dynamic control method based on an intelligent wearable device, characterized in that: The method comprises the following steps: Obtain the wearer's physiological information and location information through smart wearable devices and pre-process the data; Based on the preprocessed data, draw a historical data trend chart, extract the continuous time interval corresponding to the maximum value of the data point, select the minimum value as the selected period, and determine the data collection time interval; According to the wearer's heart rate information, the wearer's state is divided into the first and second activity states by comparing it with the set heart rate threshold; The wearer's mobile state is determined based on the wearer's information and location information. The mobile state generates information packet P1, and the stationary state generates information packet P2. Determine whether the wearer is moving or stationary. When the wearer is moving, the sound is balanced and the bass is adjusted according to the information package P1; when the wearer is stationary, the speaker sound weight is set according to the information package P2, and the audio is adjusted by building a model.

2. According to claim 1, the intelligent audio dynamic control method based on the intelligent wearable device is characterized in that: Collect sensor data of smart wearable devices in different usage scenarios and pre-process the collected sensor data. The specific steps are as follows: In the scenario of wearing a smart wearable device, the wearer's information is collected using the built-in sensor of the smart wearable device at the initial sampling time interval. The acquired wearer information includes physiological information and location information, wherein the physiological information includes heart rate information and blood oxygen information; For heart rate information collection, a heart rate sensor is used to collect data and record the heart rate value H every T seconds. For blood oxygen information collection, a blood oxygen sensor is used to collect data and record the blood oxygen saturation value O every T seconds. In terms of location information collection, the built-in GPS positioning module of the smart wearable device is used to collect location data and record the location coordinate information L every T seconds; The acquired data is preprocessed and processed by the sliding average filtering method to obtain the preprocessed heart rate data H ’ , the pre-processed blood oxygen saturation data O is obtained by using the median filter algorithm ’ , the pre-processed position coordinate data L is obtained by using the Kalman filter algorithm ’ , normalize different types of data, and normalize the heart rate, blood oxygen saturation and position coordinate data to the [0,1] interval.

3. According to claim 2, the intelligent audio dynamic control method based on the intelligent wearable device is characterized in that: The data collection interval is determined as follows: Acquire the wearer's physiological information and location information collected by the smart wearable device in the past period of time, sort each type of data in chronological order to obtain several historical data points, mark the historical data points of each type of data in a plane rectangular coordinate system, and connect adjacent historical data points in chronological order to obtain a historical data trend chart of each data type. For the historical data trend chart of each data type, extract the continuous time interval corresponding to the minimum and maximum values ​​of the data points, sort all the extracted continuous time intervals in descending order, and for each data type, select the minimum value in the continuous time interval as the selected period of the data type, and use the selected period to determine the time interval for collecting data of this type by the smart wearable device.

4. According to claim 2, the intelligent audio dynamic control method based on the intelligent wearable device is characterized in that: Classifying the wearer's state into a first-category activity state and a second-category activity state according to the collected wearer's heart rate information; The first type of activity state is defined as follows: if the collected heart rate of the wearer is greater than a set heart rate threshold, the wearer is judged to be in the first type of activity state, wherein the set heart rate threshold is calculated through historical data; The second type of activity state is defined as follows: if the collected heart rate of the wearer is less than or equal to a set heart rate threshold, it is determined that the wearer is in the second type of activity state.

5. According to claim 4, the intelligent audio dynamic control method based on the intelligent wearable device is characterized in that: The wearer's movement behavior is judged based on the acquired wearer information and location information. First, it is judged whether the wearer is in a moving state. If the wearer is in a moving state, the wearer's heart rate information in the moving state is integrated to generate an information package P1; if the wearer is in a stationary state, the wearer's heart rate information and location information in the stationary state are integrated together to generate an information package P2.

6. The intelligent audio dynamic control method based on an intelligent wearable device according to claim 5, characterized in that: When the wearer is in a moving state, first, the built-in speaker of the smart wearable device is balanced according to the information of information packet P1; Next, it is determined whether the wearer is in the first type of activity state or the second type of activity state. If the wearer is in the first type of activity state, the overall output of the speaker matrix is ​​adjusted to a soft hypnotic state by lowering the subwoofer; if the wearer is in the second type of activity state, the overall output of the speaker matrix is ​​adjusted to meet the audio style requirements of the second type of activity state by raising the subwoofer.

7. The intelligent audio dynamic control method based on an intelligent wearable device according to claim 5, characterized in that: When the wearer is in a stationary state, first, the sound weight of the built-in speaker of the smart wearable device is set according to the information in the information package P2; Next, it is determined whether the wearer is in the first type of activity state or the second type of activity state. If the wearer is in the first type of activity state, the heavy bass is lowered, and the gain of the frequency band of each speaker is fine-tuned in combination with the acoustic environment of the wearer's location to adjust the overall output of the speaker matrix to a soft hypnotic state; If the wearer is in the second type of activity, the bass is turned up, and combined with the wearer's location information, the mid- and high-frequency output of speakers in different positions is adjusted to adjust the overall output of the speaker matrix to meet the audio style requirements of the second type of activity.

8. The intelligent audio dynamic control method based on an intelligent wearable device according to claim 7, characterized in that: According to the information in information package P2, set the sound weight of the built-in speaker of the smart wearable device. The specific steps are as follows: Acoustic environment information, historical sound weights and audio effect feedback, as well as the physiological information and location information collected above, are obtained by collecting. Acoustic environment information uses the built-in microphone of the smart wearable device to collect ambient sound and analyze its frequency distribution, volume and reverberation time acoustic parameters. Historical sound weights and audio effect feedback record the sound weights set for the speakers in the past and the corresponding audio effect evaluation. The evaluation method is the user's subjective feedback, including scoring and text evaluation, and indicators obtained through objective audio analysis tools, including sound clarity and timbre balance. For heart rate data, the statistical features of the mean and standard deviation of the heart rate are extracted. For acoustic environment data, the energy distribution features of the main frequency components are extracted. Through correlation analysis, the features with a high correlation with the utterance weight are selected as the input of the model. In terms of model selection, the 5-fold cross-validation method is used to evaluate the performance of the linear regression model, decision tree model and neural network model respectively. For the linear regression model, the mean square error (MSE) is used as the evaluation indicator; for the decision tree model, the MSE evaluation is used, and the complexity of the model is taken into account at the same time. The complexity is measured by the depth of the tree. For the neural network model, the cross entropy loss function is used for evaluation. By comparing the performance of different models in cross-validation, the model with the best performance is selected, and the voice weight of the built-in speaker of the smart wearable device is set according to the voice weight value predicted by the model.

9. The intelligent audio dynamic control method based on an intelligent wearable device according to claim 7, characterized in that: The specific steps for adjusting the built-in speaker frequency band of smart wearable devices are as follows: Collect audio data of the smart wearable device's status reflected by the heart rate, blood oxygen saturation and location information of different wearers, as well as the corresponding user status information and audio adjustment records, and build a database; Get the records in the database. The built-in speaker of the smart wearable device has n adjustable frequency bands, a1, a2, ..., a n Respectively represent the adjustment amount of the 1st, 2nd, ..., nth frequency band in the current record, t1 represents the change in the evaluation index of the audio effect after adjustment, and the function equation f(a1, a2, ..., a n )=t1; Select a neural network model to generate a multivariate mapping function, and output a multivariate mapping function P(x1, x2, ..., x) between the frequency band adjustment amount and the audio effect evaluation index change amount. n ), where x1,x2,...,x n Respectively represent the adjustment amounts of the 1st, 2nd, ..., nth frequency bands; The wearer-related data collected in real time and preprocessed are mapped according to the multivariate mapping function P(x1,x2,...,x n ) input requirements are sorted and converted before inputting into the function to calculate the predicted frequency band adjustment y1, y2, ..., y n , wherein the predicted frequency band adjustment amount represents the parameter value that needs to be adjusted for each frequency band in order to achieve the best audio effect.

Images