Active noise reduction control system and method of smart phone

Abstract

The invention discloses an active noise reduction control system and method for a smart phone. The active noise reduction control system comprises an environment sound acquisition module, a motion state sensing module, a deep learning processing module, a noise reduction signal generation module, a noise reduction execution module, a user preference setting module, a scene recognition module and a power management module. The environment sound collection module collects signals in an omnibearing mode through a multi-microphone space array, data quality is guaranteed through broadband response and a high sampling rate, the motion state sensing module monitors mobile phone motion parameters in real time and provides a basis for noise reduction strategy adjustment, the deep learning processing module adopts a mixed model architecture to be combined with GPU acceleration, noise characteristics are accurately analyzed, and noise reduction is achieved. The noise reduction signal generation module can generate an adaptive noise reduction signal according to an analysis result, and the noise reduction execution module performs active feedback control and a multi-mode sounding strategy; the precision, adaptability and intelligent level of noise reduction are remarkably improved, and the auditory experience of a user using a mobile phone in different scenes is greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of noise reduction for smart phones, and particularly to an active noise reduction control system and method for smart phones. Background Art

[0002] In the current usage scenarios of smart phones, users' demand for a quiet and comfortable audio environment is increasing day by day. Whether it is making a call on a noisy street or enjoying music on the subway, environmental noise seriously affects the user experience. With the continuous improvement of people's requirements for the audio quality of mobile devices, active noise reduction technology has become the key to enhancing the user experience of smart phones. In various complex environments, such as shopping malls, airports, etc., environmental noise has the characteristics of diversity and dynamic change, and traditional audio processing methods are difficult to meet users' expectations for noise reduction effects, which makes the research and development of an active noise reduction control system and method for smart phones necessary.

[0003] Existing active noise reduction technologies have many drawbacks in practical applications. In terms of environmental sound collection, many systems use a single microphone configuration, which cannot collect sound signals omnidirectionally and multi-angularly, and has a narrow frequency response range and a low sampling rate, resulting in poor quality of the collected data and difficulty in accurately reflecting the true situation of complex environmental noise, bringing difficulties to subsequent noise reduction processing. In terms of noise reduction strategy adjustment, there is a lack of effective monitoring and utilization of the mobile state of the phone, and it is impossible to optimize the noise reduction strategy in real time according to the dynamic changes of the phone, resulting in a significant reduction in the noise reduction effect during the user's movement.

[0004] In the noise reduction signal processing link, traditional methods mostly use simple models and algorithms, which are difficult to accurately analyze the noise characteristics, and the generated noise reduction signals have poor adaptability. For example, the method of uniformly processing broadband signals cannot perform refined processing according to the characteristics of different frequency noises, and the noise reduction accuracy is insufficient. At the same time, during the noise reduction execution process, there is a lack of an effective feedback control mechanism and a flexible sound generation strategy, and it cannot be dynamically adjusted according to environmental changes and user needs, resulting in limited noise reduction effects and poor user experience. In addition, existing technologies often ignore users' personalized needs and do not fully consider the differences in different scenarios, and there are also deficiencies in operation efficiency and power management, and it is impossible to achieve an efficient and lasting noise reduction function. Summary of the Invention

[0005] In order to overcome the drawbacks and deficiencies of the existing technology, the present invention provides an active noise reduction control system and method for smart phones.

[0006] On the one hand, the present application provides an active noise reduction control system for smart phones, and the system includes:

[0007] Ambient Sound Acquisition Module: Multiple high-sensitivity microphones are set at different positions of the mobile phone to form a spatial array, which includes at least three omnidirectional microphones and two directional microphones to collect ambient sound signals omnidirectionally and from multiple angles. The microphones have broadband response characteristics, with a frequency response range of 20Hz - 20kHz and a sampling rate of not less than 48kHz;

[0008] Motion State Sensing Module: An accelerometer and a gyroscope are integrated to monitor the acceleration and angular velocity motion parameters of the mobile phone in real time. The measurement range of the accelerometer is ±16g, and the resolution reaches 0.001g; the measurement range of the gyroscope is ±2000° / s, and the resolution is 0.0625° / s. This module is connected to the main processor of the mobile phone through an I2C or SPI communication interface and transmits the collected motion data at a frequency of not less than 100Hz;

[0009] Deep Learning Processing Module: Adopts a hybrid model architecture combining a convolutional neural network and a recurrent neural network. The convolutional neural network part includes convolutional layers and pooling layers for extracting spatial features in the ambient sound signal. The convolutional kernel sizes are diverse, such as 3x3 and 5x5, and there are at least three convolutional layers, with a ReLU activation function connected after each convolutional layer; the pooling layer uses max pooling, and the pooling kernel size is 2x2. The recurrent neural network part uses long short-term memory network units with 256 memory units to process the time series features of the sound signal and capture the dynamic change rules of noise. This module uses the mobile phone's GPU to accelerate calculations and performs real-time processing and analysis on the collected sound and motion data without affecting the normal usage performance of the mobile phone.

[0010] Noise Reduction Signal Generation Module: Based on the output result of the deep learning processing module, generates a noise reduction signal with the same amplitude and opposite phase as the ambient noise. Adjusts the noise reduction signal through an adaptive filter, which uses the least mean square algorithm or the recursive least squares algorithm to update the filter coefficients in real time according to the change of the ambient noise to ensure the accuracy and stability of the noise reduction effect. At the same time, in combination with the audio power amplifier of the mobile phone, amplifies the noise reduction signal to an appropriate power level and drives the speaker to output.

[0011] Noise Reduction Execution Module: Consists of multiple micro speakers, distributed near the ear or the sound-emitting area of the mobile phone. The speakers adopt a flat design to adapt to the limited space inside the mobile phone. Their frequency response range matches that of the microphones in the ambient sound acquisition module, which is 20Hz - 20kHz, and they have good linearity, capable of accurately restoring the noise reduction signal and canceling the ambient noise. In addition, it can automatically adjust the sound emission mode and volume of the speakers according to different usage scenarios of the mobile phone, such as making calls, listening to music, watching videos, etc.

[0012] Furthermore, the ambient sound acquisition module further includes:

[0013] A barometric pressure sensor is used to detect changes in ambient barometric pressure, thereby assisting in judging the type of ambient noise. For example, the barometric pressure change during an aircraft flight is different from that of the ground environment, and based on this, the noise can be more accurately identified. The measurement range of this barometric pressure sensor is 300 hPa - 1100 hPa, the resolution is 0.1 hPa, and the measurement accuracy is ±0.3 hPa.

[0014] A temperature sensor monitors the ambient temperature because temperature affects the sound propagation characteristics and thus the noise characteristics. The measurement range of the temperature sensor is -40°C - 125°C, the accuracy is ±0.5°C, and it is connected to the main processor through the I2C interface. The collected barometric pressure and temperature data are transmitted to the deep learning processing module together with the sound signal for comprehensive analysis.

[0015] Furthermore, the deep learning processing module also has:

[0016] An online update mechanism that uses the incremental learning algorithm. When the mobile phone collects new ambient sound and motion data, instead of retraining the entire model, it performs incremental updates based on the existing model parameters. For example, the Bayesian online learning algorithm is used to continuously adjust the posterior distribution of the model parameters according to the new data, enabling the model to adapt to environmental changes in real time and improving the noise reduction effect.

[0017] A model fusion technology that fuses multiple deep learning models with different structures or trained on different data sets. For example, the weighted average fusion method is adopted. According to the noise reduction performance of each model in different scenarios, different weights are assigned to each model, and the output results of multiple models are integrated to improve the accuracy and reliability of the noise reduction decision.

[0018] Furthermore, the noise reduction signal generation module further includes:

[0019] A subband decomposition unit that divides the collected ambient sound signal and the generated noise reduction signal into multiple subbands according to frequency. For example, a uniform filter bank is used to divide the frequency range of 0 Hz - 20 kHz into 16 subbands. Noise reduction processing is performed separately in each subband, and the adaptive filter coefficients are independently adjusted according to the noise characteristics of different subbands, so as to more finely eliminate the noise in different frequency bands and improve the overall noise reduction effect.

[0020] A phase compensation circuit. Since there is a phase delay during the sound propagation process, this circuit is used to accurately compensate the phase of the noise reduction signal to ensure that the noise reduction signal is accurately aligned with the ambient noise in the time domain to achieve the best noise cancellation effect. The phase compensation range is ±180°, and the resolution is 0.1°.

[0021] Furthermore, the noise reduction execution module also has:

[0022] Active feedback control mechanism. An additional microphone is set near the speaker to collect in real time the mixed signal after the noise reduction signal output by the speaker is superimposed with the residual ambient noise. This mixed signal is fed back to the noise reduction signal generation module for comparative analysis with the original ambient sound signal, further adjusting the amplitude and phase of the noise reduction signal to form a closed-loop control and continuously optimize the noise reduction effect.

[0023] Multi-mode sound generation strategy. Different sound generation modes can be switched according to the usage scenario of the mobile phone and the user's needs. For example, in a call scenario, a mono sound generation mode is adopted to concentrate on canceling the noise in the call direction; in a music playback scenario, it is switched to a stereo sound generation mode to ensure the stereo effect of the music while reducing noise; in a noisy outdoor environment, the speaker volume is automatically increased and the low-frequency noise reduction effect is enhanced.

[0024] Furthermore, the system further includes:

[0025] User preference setting module. The user can perform personalized settings on the active noise reduction system through the mobile phone application. For example, the user can select different noise reduction intensity levels, such as mild, moderate, and severe noise reduction; and can also customize the noise reduction priority for certain specific frequency noises. The system will adjust the parameters of the deep learning processing module and the noise reduction signal generation module according to the user settings to meet the user's personalized needs.

[0026] Scene recognition module. Using machine learning algorithms, based on the data collected by the ambient sound collection module, the motion state perception module, and other sensors (such as barometric pressure and temperature sensors), it identifies the environmental scene where the mobile phone is located, such as an airport, subway, office, indoor, etc. For different scenes, different noise reduction parameter configuration files are pre-stored, and the system automatically calls the corresponding configuration files to optimize the noise reduction effect.

[0027] Furthermore, the system adopts a distributed computing architecture:

[0028] Some computing tasks are undertaken by the local processor of the mobile phone, such as the preliminary processing of the data of the ambient sound collection module, the data collection and transmission of the motion state perception module, etc. The local processor uses a high-performance mobile CPU, combined with GPU acceleration, to ensure that tasks with high real-time requirements can be executed quickly.

[0029] Complex deep learning model computing tasks, such as the operations of convolutional neural networks and recurrent neural networks, can be transmitted to the cloud server through a wireless network (such as Wi-Fi or 5G) for processing. The cloud server has powerful computing resources and can quickly complete the operations of the deep learning model and return the processing results to the mobile phone for generating the noise reduction signal. This distributed computing architecture can not only make full use of the local resources of the mobile phone but also leverage the powerful computing power of the cloud to improve the overall performance of the system.

[0030] Furthermore, the system further includes:

[0031] A power management module, which adopts a dynamic power management strategy according to the power consumption characteristics of each module of the active noise cancellation system. For example, when the ambient noise is low or the mobile phone is in a stationary state, it reduces the operation frequency of the deep learning processing module and reduces the power consumption of the GPU; when the amount of data collected by the microphone and sensors is small, it reduces their sampling frequency to save power. At the same time, it optimizes the power output of the execution modules such as the speaker, dynamically adjusts the output power according to the noise cancellation requirements, and minimizes the overall power consumption of the system while ensuring the noise cancellation effect, thereby extending the battery life of the mobile phone.

[0032] On the other hand, the present application also provides an active noise cancellation control method for a smart phone, which includes the following steps:

[0033] Data acquisition step: Through the ambient sound acquisition module, the motion state perception module, and other auxiliary sensors (such as barometric pressure and temperature sensors), synchronously acquire ambient sound signals, mobile phone motion data, and environmental parameter data at a set frequency. The ambient sound acquisition module acquires sound signals at a sampling rate not lower than 48 kHz, the motion state perception module acquires motion data at a frequency not lower than 100 Hz, and other sensors acquire data at their respective optimal response frequencies, and transmit these data to the main processor of the mobile phone through the corresponding communication interfaces.

[0034] Data preprocessing step: Preprocess the collected raw data, including operations such as removing noise interference, data calibration, and normalization. For ambient sound signals, digital filtering technology is used to remove high-frequency or low-frequency noise; the motion data is calibrated to eliminate sensor drift errors; all data is normalized to make it within the same numerical range, facilitating subsequent processing by the deep learning model.

[0035] Deep learning processing step: Input the preprocessed data into the deep learning processing module, and use a hybrid model combining a convolutional neural network and a recurrent neural network to extract the spatial and temporal features in the data, and analyze the type, intensity, and change trend of the ambient noise. The model continuously learns and trains the data, optimizes the model parameters, and improves the ability to identify and predict noise.

[0036] Noise cancellation signal generation step: Generate a noise cancellation signal with the same amplitude and opposite phase as the ambient noise according to the output result of the deep learning processing module. Using an adaptive filter, according to the real-time change of the ambient noise, update the filter coefficients using the least mean square algorithm or the recursive least squares algorithm, and adjust the noise cancellation signal to make it more accurately match the ambient noise.

[0037] Noise reduction execution steps: Amplify the generated noise reduction signal through an audio power amplifier to an appropriate power level, drive the miniature speaker of the noise reduction execution module to output, and cancel the ambient noise. At the same time, according to the active feedback control mechanism, the mixed signal after the speaker output is collected in real time and fed back to the noise reduction signal generation step to further optimize and adjust the noise reduction signal.

[0038] Furthermore, the method further includes the following steps:

[0039] User preference application steps: Read the personalized settings of the user in the user preference setting module, such as the noise reduction intensity level, the noise reduction priority of specific frequencies, etc. According to the user settings, adjust the relevant parameters in the deep learning processing step and the noise reduction signal generation step to make the active noise reduction system meet the user's personalized needs.

[0040] Scene recognition application steps: Use the scene recognition module to identify the environmental scene where the mobile phone is located. According to the recognition result, call the corresponding configuration from the pre-stored noise reduction parameter configuration files for different scenes to optimize and adjust the parameters in the entire active noise reduction control process to meet the noise reduction requirements in different environmental scenes.

[0041] Beneficial effects:

[0042] The present invention proposes an active noise reduction control system and method for a smart phone. In terms of the system, multiple modules cooperate. The environmental sound collection module collects signals omnidirectionally through a multi-microphone spatial array, and the broadband response and high sampling rate ensure data quality. The motion state perception module monitors the motion parameters of the mobile phone in real time to provide a basis for adjusting the noise reduction strategy. The deep learning processing module uses a hybrid model architecture combined with GPU acceleration to accurately analyze the noise characteristics. The noise reduction signal generation module can generate an adaptive noise reduction signal according to the analysis result, and the sub-band decomposition and phase compensation improve the noise reduction accuracy. The active feedback control and multi-mode sound generation strategy of the noise reduction execution module effectively enhance the noise reduction effect and user experience. At the same time, the system is also equipped with modules such as user preference settings, scene recognition, distributed computing, and power management to meet personalized needs and improve the operation efficiency and battery life. At the method level, from data collection, preprocessing, to using a deep learning model to analyze, generate and adjust the noise reduction signal, and then to optimize according to the feedback mechanism and adapt the parameters in combination with user preferences and scene recognition, a complete and efficient process is formed. Overall, this application significantly improves the accuracy, adaptability and intelligent level of noise reduction, and greatly enhances the auditory experience of users when using the mobile phone in different scenarios. Description of the drawings

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0044] Figure 1 It is a block diagram of the module composition of the system of the present invention;

[0045] Figure 2 It is a flowchart of the operation of the method of the present invention. Specific embodiments

[0046] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will further describe this application in detail with reference to the drawings and specific embodiments.

[0047] As Figure 1 shown, on the one hand, this embodiment provides an active noise reduction control system for a smart phone, which is characterized in that the system includes:

[0048] Ambient sound acquisition module: A number of high-sensitivity microphones are arranged at different positions of the mobile phone to form a spatial array, including three omnidirectional microphones and two directional microphones, to collect ambient sound signals in all directions and at multiple angles. The microphones have broadband response characteristics, with a frequency response range of 20 Hz - 20 kHz, and a sampling rate of not less than 48 kHz;

[0049] Specifically, the environmental sound collection module is the data source of the active noise cancellation control system. Its ability to collect environmental sound signals in all directions and from multiple angles lays the foundation for subsequent accurate analysis and elimination of environmental noise. By setting up multiple microphones at different positions on the mobile phone to form a spatial array, it can capture sounds from all directions, ensuring a comprehensive perception of complex environmental noise. This module uses a spatial array composed of three omnidirectional microphones and two directional microphones. Omnidirectional microphones can receive sounds from all directions, while directional microphones can specifically collect sounds in a particular direction. The combination of the two can achieve more detailed collection of environmental sounds. The microphones have broadband response characteristics, with a frequency response range of 20 Hz - 20 kHz. This range covers the main frequency range of human hearing and can completely collect various sounds in the environment. The sampling rate is not less than 48 kHz. A high sampling rate can ensure that the collected sound signals are more accurate and detailed, reducing information loss. In various complex environments, this module can play an important role. For example, on the street, it can collect the driving sounds of cars, the noises of the crowd, etc.; in the shopping mall, it can capture the promotional broadcasts of merchants, the conversations of people, etc.; in the subway, it can also collect the noises of train operation, the footsteps of passengers, etc. These rich sound signals will be transmitted to the subsequent modules for processing.

[0050] Motion state perception module: Integrating an accelerometer and a gyroscope, it monitors the acceleration and angular velocity motion parameters of the mobile phone in real time. The measurement range of the accelerometer is ±16 g, and the resolution reaches 0.001 g; the measurement range of the gyroscope is ±2000 ° / s, and the resolution is 0.0625 ° / s. This module is connected to the main processor of the mobile phone through the I2C or SPI communication interface and transmits the collected motion data at a frequency not less than 100 Hz;

[0051] Specifically, during the use of the mobile phone, it is often in different motion states, and the changes in the motion state will affect the environmental sound and the noise cancellation effect. The motion state perception module monitors the acceleration and angular velocity motion parameters of the mobile phone in real time, enabling the system to dynamically adjust the noise cancellation strategy according to the motion situation of the mobile phone, improving the accuracy and adaptability of noise cancellation.

[0052] Model: This module integrates an accelerometer and a gyroscope. The accelerometer is used to measure the acceleration of the mobile phone, and the gyroscope is used to measure the angular velocity of the mobile phone. By the collaborative work of these two sensors, the motion state information of the mobile phone can be obtained comprehensively and accurately. The measurement range of the accelerometer is ±16g, and the resolution reaches 0.001g, which can accurately measure the motion of the mobile phone under different accelerations; the measurement range of the gyroscope is ±2000° / s, and the resolution is 0.0625° / s, which can accurately sense the rotation speed of the mobile phone. This module is connected to the main processor of the mobile phone through the I2C or SPI communication interface and transmits the collected motion data at a frequency not lower than 100Hz to ensure the real-time nature of the data. When the user holds the mobile phone and walks, the accelerometer and gyroscope will detect the shaking and movement of the mobile phone, and the system can adjust the noise reduction strategy according to these motion data to avoid the additional noise generated by the motion from affecting the noise reduction effect; in scenarios such as taking a car or running, the noise reduction parameters can also be dynamically optimized according to the motion state of the mobile phone to provide a more stable noise reduction experience.

[0053] Deep learning processing module: It adopts a hybrid model architecture combining a convolutional neural network and a recurrent neural network. The convolutional neural network part includes convolutional layers and pooling layers, which are used to extract the spatial features in the environmental sound signal. Each convolutional layer is followed by a ReLU activation function; the pooling layer uses the maximum pooling method, and the pooling kernel size is 2x2. The recurrent neural network part uses long short-term memory network units with 256 memory units, which are used to process the time series features of the sound signal and capture the dynamic change law of the noise. This module uses the mobile phone GPU to accelerate the calculation and, without affecting the normal use performance of the mobile phone, performs real-time processing and analysis on the collected sound and motion data;

[0054] Specifically, the deep learning processing module is the core of the entire active noise reduction control system. It is responsible for deeply analyzing and processing the collected sound and motion data. By extracting the spatial features and time series features in the environmental sound signal, it can accurately capture the dynamic change law of the noise and provide a basis for generating effective noise reduction signals in the follow-up.

[0055] Model: This module adopts a hybrid model architecture that combines a Convolutional Neural Network (CNN) and a Recurrent Neural Network (RNN). The convolutional neural network part includes convolutional layers and pooling layers. The convolutional layers extract spatial features from the environmental sound signals through convolutional kernels. After each convolutional layer, a ReLU activation function is connected, which can introduce non-linear factors and enhance the expressive ability of the model. The pooling layer uses the maximum pooling method with a pooling kernel size of 2x2 to reduce the dimension of the feature map and reduce the computational amount. The recurrent neural network part uses Long Short-Term Memory (LSTM) units with 256 memory units, which can process the time series features of the sound signals and capture the dynamic change rules of the noise. Using the mobile phone GPU to accelerate the calculation, on the premise of not affecting the normal use performance of the mobile phone, real-time processing and analysis of the collected data are realized. Through the parallel computing ability of the GPU, the calculation speed of the model can be greatly improved, ensuring that the system can respond to the changes in environmental noise in a timely manner. In scenarios with complex and variable environmental noise, such as construction sites and airport waiting halls, the deep learning processing module can quickly and accurately analyze the type, intensity, and change trend of the noise. For example, at a construction site, there may be various noises such as the roar of machines and the shouts of workers at the same time. This module can identify the characteristics of different types of noises through feature extraction and analysis of the sound signals, providing accurate information for subsequent noise reduction processing.

[0056] Noise reduction signal generation module: Based on the output results of the deep learning processing module, a noise reduction signal with the same amplitude and opposite phase as the environmental noise is generated. The noise reduction signal is adjusted through an adaptive filter, which uses the least mean square algorithm or the recursive least squares algorithm to update the filter coefficients in real time according to the changes in the environmental noise. At the same time, combined with the audio power amplifier of the mobile phone, the noise reduction signal is amplified to an appropriate power level to drive the speaker to output;

[0057] Specifically, the main function of the noise reduction signal generation module is to generate a noise reduction signal with the same amplitude and opposite phase as the ambient noise according to the output result of the deep learning processing module. In this way, when the noise reduction signal is superimposed on the ambient noise, they can cancel each other out, thereby achieving the purpose of reducing the ambient noise. Based on the output result of the deep learning processing module, this module uses an adaptive filter to adjust the noise reduction signal. The adaptive filter adopts the least mean square algorithm (LMS) or the recursive least squares algorithm (RLS), which can update the filter coefficients in real time according to the change of the ambient noise, so that the noise reduction signal can better adapt to different noise environments. Combined with the audio power amplifier of the mobile phone, the noise reduction signal is amplified to an appropriate power level to drive the speaker to output. The role of the power amplifier is to enhance the intensity of the noise reduction signal to ensure that the speaker can output a noise reduction signal with sufficient intensity to cancel the ambient noise. In various scenarios that require noise reduction, this module can play a key role. For example, during a call, it can generate a corresponding noise reduction signal according to the ambient noise situation around, reduce the interference of external noise on the call, and improve the call quality; when listening to music, it can effectively reduce the ambient noise and allow users to enjoy music more attentively.

[0058] Noise reduction execution module: It consists of multiple micro speakers, which are distributed near the ear or the sound - emitting area of the mobile phone. The speakers adopt a flat design to adapt to the limited space inside the mobile phone. Their frequency response range matches that of the microphone of the ambient sound acquisition module, which is 20Hz - 20kHz, and automatically adjusts the sound - emitting mode and volume of the speakers according to different usage scenarios of the mobile phone.

[0059] Specifically, the noise reduction execution module is a key link to convert the noise reduction signal into an actual noise reduction effect. It outputs the noise reduction signal through multiple micro speakers to cancel the ambient noise with it, thereby achieving the purpose of reducing the ambient noise. This module consists of multiple micro speakers, which are distributed near the ear or the sound - emitting area of the mobile phone. The speakers adopt a flat design, which can adapt to the limited space inside the mobile phone while ensuring good sound output effect. The frequency response range of the speakers matches that of the microphone of the ambient sound acquisition module, which is 20Hz - 20kHz, ensuring that the noise reduction signal corresponding to the ambient noise can be accurately output. According to different usage scenarios of the mobile phone, this module can automatically adjust the sound - emitting mode and volume of the speakers to achieve the best noise reduction effect. In different usage scenarios, the noise reduction execution module will automatically adjust the working mode. For example, in the call scenario, it will adopt the mono - sound - emitting mode to focus on canceling the noise in the call direction; in the music - playing scenario, it will switch to the stereo - sound - emitting mode to ensure the stereo effect of the music while reducing noise; in a noisy outdoor environment, it will automatically increase the speaker volume and enhance the low - frequency noise reduction effect to effectively reduce the interference of external noise.

[0060] Preferably, the ambient sound acquisition module further includes:

[0061] A barometric pressure sensor, used to detect changes in ambient barometric pressure and assist in judging the type of ambient noise. The measurement range of this barometric pressure sensor is 300 hPa - 1100 hPa, the resolution is 0.1 hPa, and the measurement accuracy is ±0.3 hPa;

[0062] Specifically, in the active noise cancellation control system of a smartphone, the barometric pressure sensor is of great significance. Changes in ambient barometric pressure can affect sound propagation. In different barometric pressure environments, the propagation characteristics, attenuation degrees, etc. of sound are different, which in turn leads to different characteristics of ambient noise. By detecting changes in ambient barometric pressure, the system can assist in judging the type of ambient noise. For example, at high altitudes, the barometric pressure is low, and the characteristics of noises such as wind noise are different from those at low altitudes; in a closed space, the barometric pressure is relatively stable, while in scenarios such as elevators with rapid ascent and descent, the barometric pressure will change significantly, and the corresponding noises will also be different. Therefore, the barometric pressure sensor provides additional environmental information for the system, helping to analyze and process ambient noise more accurately. The barometric pressure sensor here is a device that can convert the physical quantity of ambient barometric pressure into a measurable electrical signal. It senses the change in barometric pressure through internal pressure-sensitive elements, such as piezoresistive, capacitive, etc., and converts it into an electrical signal for output. The measurement range is 300 hPa - 1100 hPa, and this range covers the common barometric pressure values in most regions of the earth, enabling effective measurement of barometric pressure from low pressure in high-altitude areas to high pressure in low-altitude areas. The resolution is 0.1 hPa, meaning it can accurately detect minute changes in barometric pressure, providing more detailed environmental information for the system. The measurement accuracy is ±0.3 hPa, ensuring the reliability of the measurement results, enabling the system to judge the type of noise based on accurate barometric pressure data. The pressure-sensitive element inside the barometric pressure sensor senses the change in ambient barometric pressure and converts it into an electrical signal. This electrical signal is processed by an internal signal conditioning circuit for amplification, filtering, etc., and then the processed digital signal is transmitted to the main processor of the mobile phone through a specific interface (such as I2C or SPI, etc.). The main processor then transmits these barometric pressure data together with other data such as sound signals to a deep learning processing module for comprehensive analysis. When the user is in different geographical environments or scenarios, the barometric pressure sensor can play a role. For example, during mountain climbing, as the altitude increases, the barometric pressure gradually decreases. After the barometric pressure sensor detects the change in barometric pressure, the system can combine the sound signal to judge the characteristics of noises such as wind noise at this time and adjust the noise cancellation strategy. When taking a plane, the barometric pressure changes drastically during takeoff and landing of the plane, and the sensor can capture these changes in a timely manner to help the system adapt to the ambient noise at different stages. During the elevator's ascent and descent, the rapid change in barometric pressure can also be detected by the sensor, enabling the system to process the noise generated by the elevator operation more accurately.

[0063] A temperature sensor monitors the ambient temperature. The measurement range of the temperature sensor is -40°C to 125°C, with an accuracy of ±0.5°C. It is connected to the main processor through the I2C interface and transmits the collected air pressure and temperature data, along with the sound signal, to the deep learning processing module for comprehensive analysis.

[0064] Specifically, temperature also affects the propagation of sound and the characteristics of noise. In different temperature environments, the density of air, the speed of sound, etc. will change, thus altering the laws of sound propagation and attenuation and affecting the characteristics of environmental noise. By monitoring the ambient temperature, the system can comprehensively consider the impact of temperature factors on noise and further improve the analysis and processing capabilities of environmental noise. For example, in hot summers and cold winters, the same environmental noise may exhibit different characteristics due to temperature differences. A temperature sensor is a device that can sense the ambient temperature and convert it into an electrical signal. Common types of temperature sensors include thermocouples, thermistors, etc., which utilize the temperature characteristics of materials to measure temperature. The measurement range is -40°C to 125°C, which covers the temperature ranges in most natural environments and daily usage scenarios. The accuracy of ±0.5°C ensures the accuracy of the measurement results, enabling the system to perform noise analysis and processing based on relatively accurate temperature data. The temperature sensor senses the change in ambient temperature through its internal sensitive element and converts the temperature signal into an electrical signal. This electrical signal is processed by the signal conditioning circuit and then transmitted to the main processor of the mobile phone through the I2C interface. The main processor transmits the collected temperature data, along with the sound signal, air pressure data, etc., to the deep learning processing module, where comprehensive analysis is carried out to consider the impact of temperature factors on environmental noise. In different seasons and environments, the temperature sensor can play an important role. In a high-temperature environment in summer, the speed of sound propagation may increase, and the propagation characteristics of noise will also change. After the temperature sensor detects the high temperature, the system can adjust the noise reduction strategy in combination with other data to adapt to the noise characteristics in the high-temperature environment. In cold winters, the temperature is low, the air density is high, and sound propagation may be affected to a certain extent. The sensor enables the system to promptly sense the temperature change and optimize the noise reduction effect. In addition, in some special scenarios, such as when near heat-generating devices (such as computers, engines, etc.), the temperature will rise, and the temperature sensor can detect this local temperature change and help the system process the surrounding noise more accurately.

[0065] Preferably, the deep learning processing module further has:

[0066] An online update mechanism that uses the incremental learning algorithm. When the mobile phone collects new environmental sound and motion data, instead of retraining the entire model, it performs incremental updates based on the existing model parameters;

[0067] Specifically, in the active noise cancellation control system of a smartphone, the environment is complex and variable, and new noise scenarios will constantly emerge. Traditional model training methods often require retraining the entire model when encountering new data, which is time-consuming, consumes a large amount of computing resources, and may also affect the normal use of the phone. The online update mechanism uses incremental learning algorithms. When the phone collects new environmental sound and motion data, it can perform incremental updates based on the existing model parameters without retraining the entire model. This enables the system to quickly adapt to the new noise environment, timely adjust the noise cancellation strategy, improve the noise cancellation effect, while saving computing resources and time, and ensuring the normal performance of the phone. This mechanism is based on incremental learning algorithms. Incremental learning is a machine learning method that allows the model to continuously learn while receiving new data, rather than learning all data at once. In this scenario, when new environmental sound and motion data arrive, the model will fine-tune the existing parameters according to these new data, rather than completely discarding the old model and retraining. There is no specific module parameter mentioned here, but the performance of the incremental learning algorithm may be affected by parameters such as the learning rate. The learning rate controls the magnitude of parameter adjustment in each update of the model. A suitable learning rate can enable the model to quickly adapt to new data while avoiding over-adjustment that may cause the model to be unstable. After the phone collects new environmental sound and motion data, these data will be transmitted to the deep learning processing module. The module first preprocesses the new data to make it compatible with the data format of the existing model. Then, the incremental learning algorithm will calculate the impact of the new data on the model parameters and update them based on certain rules (such as the gradient descent method) on the basis of the existing model parameters. The updated model parameters will be used for subsequent noise cancellation processing. For example, when a user travels from one city to another, the environmental noise in different cities may vary, such as different traffic noise patterns and characteristics of crowd noise. The online update mechanism can allow the phone system to quickly update the noise cancellation model according to the newly collected environmental data after arriving in the new city to adapt to the new noise environment. Another example is that as the seasons change, the environmental noise also varies. There may be more cicada sounds in summer and more obvious wind sounds in winter. This mechanism can enable the system to adjust in a timely manner to ensure a good noise cancellation effect.

[0068] Model fusion technology integrates multiple deep learning models with different structures or trained on different datasets. Based on the noise reduction performance of each model in different scenarios, different weights are assigned to each model, and the output results of multiple models are combined. Deep learning models with different structures or trained on different datasets may have different advantages in different scenarios. A single model often struggles to achieve the best noise reduction effect in all scenarios. Model fusion technology combines multiple such models, assigns different weights to them according to their noise reduction performance in different scenarios, and combines the output results of multiple models. This can fully leverage the advantages of each model, improve the overall noise reduction performance of the system in various complex scenarios, and enhance the robustness and adaptability of the system. This technology involves multiple deep learning models with different structures or trained on different datasets, such as convolutional neural networks (CNNs) or recurrent neural networks (RNNs) with different numbers of layers and different numbers of neurons. These models can be trained on different environmental noise datasets. For example, some models are trained on urban traffic noise datasets, while others are trained on shopping mall noise datasets. The main module parameter is the weight assigned to each model. The weight assignment needs to be determined based on the noise reduction performance of each model in different scenarios. For example, in a specific scenario, if Model A has a better noise reduction effect, its weight in this scenario will be relatively higher; while if Model B has a worse noise reduction effect, its weight will be lower. First, multiple different deep learning models are integrated into the deep learning processing module. When processing environmental sound and motion data, each model processes the data and outputs its respective noise reduction results. Then, the system evaluates the performance of each model in the current scenario according to preset evaluation metrics (such as the evaluation function of the noise reduction effect). Based on the evaluation results, corresponding weights are assigned to each model. Finally, the output results of each model are weighted and summed according to the assigned weights to obtain the final noise reduction result. In a complex and changing environment, model fusion technology can play an important role. For example, in a city square scenario with both traffic noise and crowd noise, different models may have different processing capabilities for traffic noise and crowd noise. Through model fusion technology, higher weights can be assigned to the model that is more effective in processing traffic noise according to the actual situation, and appropriate weights can also be assigned to the model that is more effective in processing crowd noise, so as to combine the advantages of each model and achieve a better noise reduction effect. Another example is on an airplane, where there are both low-frequency engine noise and high-frequency noise such as conversations in the cabin. Model fusion technology can enable the system to better handle such a complex noise environment.

[0069] Preferably, the noise reduction signal generation module further includes:

[0070] A subband decomposition unit divides the collected environmental sound signal and the generated noise reduction signal into multiple subbands according to frequency, divides the frequency range of 0 Hz - 20 kHz into 16 subbands, performs noise reduction processing separately within each subband, and independently adjusts the adaptive filter coefficients according to the noise characteristics of different subbands;

[0071] Specifically, environmental noise usually contains rich frequency components, and noises of different frequencies have different characteristics and sources. The subband decomposition unit divides the collected environmental sound signal and the generated noise reduction signal into multiple subbands according to frequency, and can perform refined processing on the noise characteristics of each subband. This method avoids the limitations of unified processing of the entire broadband signal, and significantly improves the pertinence and effectiveness of noise reduction. By independently adjusting the adaptive filter coefficients within each subband, the system can more accurately match the changes of noises of different frequencies, thereby achieving a better noise reduction effect. This unit adopts a frequency-domain decomposition model, and divides the audio signal spectrum from 0 Hz to 20 kHz into multiple subbands based on mathematical methods such as Fourier transform. In this design, it is selected to be divided into 16 subbands, and each subband covers a certain frequency range. This division method achieves a good balance between computational complexity and noise reduction effect, ensuring both the fine processing of noises of different frequencies and not making the computational amount too large to affect the real-time performance of the system. The main parameters are the number of subband divisions (16 subbands) and the corresponding frequency range of each subband. The frequency range from 0 Hz to 20 kHz is evenly divided into 16 subbands, and the frequency span of each subband is approximately 1.25 kHz (20 kHz ÷ 16). In practical applications, this uniform division method can better adapt to the spectral distributions of most common noises. However, for some noise scenarios with specific frequency characteristics, non-uniform division can also be performed according to needs to further optimize the noise reduction performance. First, the original sound signal collected by the environmental sound acquisition module enters the subband decomposition unit. Here, the signal is transformed from the time domain to the frequency domain through Fourier transform, and then divided into 16 subbands according to the predetermined frequency range. For each subband, its noise characteristics, such as the intensity and frequency distribution of the noise, are calculated separately. Based on these characteristics, the coefficients of the adaptive filter are independently adjusted so that the filter can generate a noise reduction signal with the same amplitude and opposite phase as the noise in this subband. Finally, the noise reduction signals processed by each subband are transformed back to the time domain through inverse Fourier transform and synthesized to obtain the complete noise reduction signal output. In various complex noise environments, the subband decomposition unit can demonstrate powerful advantages. For example, in a factory workshop, the environmental noise includes low-frequency roars generated by machine operation, high-frequency sharp sounds generated by tool collisions, and medium-frequency noises caused by equipment vibration, etc. Through subband decomposition, the system can process the noises in different frequency bands separately and effectively reduce the interference of various noises. In the subway carriage, there are both low-frequency noises of train operation and high-frequency sounds such as broadcasts and passenger conversations. The subband decomposition unit can accurately perform noise reduction processing on the noises in different frequency ranges and improve the auditory comfort of passengers.

[0072] A phase compensation circuit, which is used to accurately compensate the phase of the noise reduction signal for precise alignment of the noise reduction signal and the environmental noise in the time domain. The phase compensation range is ±180°, and the resolution is 0.1°.

[0073] Specifically, the precise alignment of the noise reduction signal and the ambient noise in the time domain is crucial for achieving effective noise reduction. Even if the amplitude of the noise reduction signal is equal to that of the ambient noise, if there is a phase deviation, they cannot completely cancel each other out when superimposed, thus affecting the noise reduction effect. The phase compensation circuit precisely compensates the phase of the noise reduction signal to ensure the precise alignment of the noise reduction signal and the ambient noise in the time domain, enabling them to cancel each other out to the greatest extent, thereby significantly improving the performance of the noise reduction system. The phase compensation circuit adopts a phase adjustment model to adjust the phase of the noise reduction signal through a specific circuit structure or digital signal processing algorithm. Its core principle is to calculate the phase value to be compensated based on the phase difference between the ambient noise and the noise reduction signal, and then adjust the phase of the noise reduction signal through corresponding circuit components or algorithm operations. The key parameters include the phase compensation range (±180°) and the resolution (0.1°). The phase compensation range of ±180° ensures that the circuit can handle various possible phase deviation situations, and can effectively compensate for both leading and lagging phase differences. The resolution of 0.1° guarantees the accuracy of phase compensation, enabling fine adjustment of the phase to meet the strict requirements for phase alignment in different noise environments. During the generation of the noise reduction signal, the phase compensation circuit continuously monitors the phase difference between the ambient noise signal and the generated noise reduction signal. Through a dedicated phase detection algorithm, the current phase deviation value is calculated. Based on this deviation value, the circuit uses internal phase adjustment components (such as variable delay lines, digital phase shifters, etc.) to adjust the phase of the noise reduction signal. The adjusted noise reduction signal is as precisely aligned as possible with the ambient noise in the time domain, and then is output to cancel the ambient noise. Throughout the process, phase detection and compensation are dynamic and real-time processes to adapt to the continuously changing characteristics of the ambient noise. In practical applications, the phase compensation circuit plays a key role in various scenarios. For example, in an indoor environment, due to factors such as sound reflection, there may be a complex phase relationship between the ambient noise and the sound signal collected by the mobile phone. The phase compensation circuit can precisely adjust the phase of the noise reduction signal so that it can also effectively cancel the reflected ambient noise, improving the indoor noise reduction effect. In a noisy outdoor environment, such as on the street, the noise coming from different directions may have different phase changes during propagation. The phase compensation circuit can dynamically adjust the phase of the noise reduction signal according to the real-time detected phase difference to ensure effective noise reduction in various complex situations.

[0074] Preferably, the noise reduction execution module further includes:

[0075] An active feedback control mechanism sets an additional microphone near the speaker to collect in real time the mixed signal after the noise reduction signal output by the speaker is superimposed with the residual ambient noise, and feeds back this mixed signal to the noise reduction signal generation module for comparative analysis with the original ambient sound signal, further adjusting the amplitude and phase of the noise reduction signal to form a closed-loop control;

[0076] A multi-mode sound generation strategy switches different sound generation modes according to the usage scenarios of the mobile phone and user needs. In the call scenario, a mono sound generation mode is adopted to focus on canceling the noise in the call direction; in the music playback scenario, it switches to a stereo sound generation mode to ensure the stereo effect of the music while reducing noise; in a noisy outdoor environment, the speaker volume is automatically increased and the low-frequency noise reduction effect is enhanced.

[0077] Specifically, in the active noise reduction control system of a smart phone, these two characteristics of the noise reduction execution module greatly improve the noise reduction effect and user experience.

[0078] The active feedback control mechanism is of great significance. It adds an additional microphone near the speaker to collect in real time the mixed signal after the noise reduction signal output by the speaker is superimposed with the residual ambient noise. This mixed signal is fed back to the noise reduction signal generation module for comparative analysis with the original ambient sound signal. The implementation process is as follows: the additional microphone continuously collects and transmits the mixed signal, and the noise reduction signal generation module compares it with the original ambient sound signal, calculates the difference between the two, and further adjusts the amplitude and phase of the noise reduction signal accordingly to form a closed-loop control. In actual usage scenarios, such as on the subway where the ambient noise changes continuously, the active feedback control mechanism can continuously optimize the noise reduction signal, continuously reduce noise interference, and ensure a good noise reduction effect at all times.

[0079] The multi-mode sound generation strategy is also practical. It flexibly switches the sound generation mode according to the usage scenarios of the mobile phone and user needs. When implemented, the system analyzes information such as the data of the mobile phone sensors and user operations to determine the current scenario. In the call scenario, the system automatically switches to the mono sound generation mode to focus on canceling the noise in the call direction. For example, when making a call on a noisy street, it can effectively reduce the interference of surrounding noise on the call sound and improve the call clarity; in the music playback scenario, it switches to the stereo sound generation mode to both reduce the ambient noise and ensure the stereo effect of the music, allowing users to immerse themselves in the music; in a noisy outdoor environment, the system automatically increases the speaker volume and enhances the low-frequency noise reduction effect. For example, near a construction site, it can effectively shield the low-frequency machine roar and guarantee the user's auditory experience.

[0080] Preferably, the system further includes:

[0081] User preference setting module: Users can perform personalized settings on the active noise cancellation system through a mobile application. Users can select different noise cancellation intensity levels, such as mild, moderate, and severe noise cancellation, and customize the noise cancellation priority for certain specific frequency noises. The system will adjust the parameters of the deep learning processing module and the noise cancellation signal generation module according to the user settings to meet the personalized needs of users.

[0082] Scene recognition module: Using machine learning algorithms, based on the data collected by the environmental sound collection module, the motion state perception module, and other sensors, it identifies the environmental scene where the mobile phone is located. For different scenes, different noise cancellation parameter configuration files are pre-stored, and the system automatically calls the corresponding configuration file.

[0083] Specifically, in this smartphone active noise cancellation control system, the user preference setting module and the scene recognition module greatly improve the personalization and intelligence level of the system.

[0084] The user preference setting module is of great significance as it gives users the right to customize the active noise cancellation experience independently. Users can easily select different noise cancellation intensity levels such as mild, moderate, and severe through a mobile application, and can also customize the noise cancellation priority for certain specific frequency noises. The implementation process is that after the user completes the settings in the application, this setting information will be transmitted to the system, and the system will accurately adjust the parameters of the deep learning processing module and the noise cancellation signal generation module according to the user settings to fit the personalized needs of users. For example, if some users are extremely sensitive to high-frequency sharp noises, they can increase the noise cancellation priority for high-frequency specific frequency noises in the settings, and the system will focus on processing such noises to create a more comfortable auditory environment for users.

[0085] The scene recognition module is equally crucial. It uses machine learning algorithms to integrate the data collected by the environmental sound collection module, the motion state perception module, and other sensors to accurately identify the environmental scene where the mobile phone is located. The implementation process is that the system continuously collects various sensor data, and the machine learning algorithm analyzes, compares, and performs pattern recognition on these data to determine the current environmental scene. For different scenes, the system pre-stores corresponding noise cancellation parameter configuration files. Once the scene is recognized, the corresponding configuration file will be automatically called. In an actual usage scenario, when the user walks from a quiet library to a noisy street, the scene recognition module quickly recognizes the scene change and automatically calls the noise cancellation parameter configuration file suitable for the noisy street environment. The system adjusts the noise cancellation strategy accordingly, increasing the noise cancellation intensity, optimizing the noise cancellation frequency range, etc., to ensure the best noise cancellation effect in different scenes and improve the user experience.

[0086] Preferably, the system adopts a distributed computing architecture:

[0087] Part of the computing tasks are undertaken by the local processor of the mobile phone. The local processor uses a high-performance mobile CPU and combines GPU acceleration to quickly execute tasks with high real-time requirements;

[0088] For complex deep learning model computing tasks, operations of convolutional neural networks and recurrent neural networks are transmitted to the cloud server via wireless network for processing, and the processing results are returned to the mobile phone to generate noise reduction signals.

[0089] Specifically, the distributed computing architecture adopted by this smartphone active noise reduction control system greatly optimizes the system performance and has outstanding performance in terms of significance, implementation process, and usage scenarios.

[0090] In terms of significance, the distributed computing architecture makes full use of the advantages of the local processor of the mobile phone and the cloud server. The local processor of the mobile phone uses a high-performance mobile CPU combined with GPU acceleration, which can quickly execute tasks with high real-time requirements, ensuring the timely processing of data such as environmental sound collection and motion state perception by the system, and guaranteeing the real-time nature of noise reduction. Transmitting complex deep learning model computing tasks, such as operations of convolutional neural networks and recurrent neural networks, to the cloud server via wireless network for processing reduces the local computing burden of the mobile phone and utilizes the powerful computing resources of the cloud to improve the operation efficiency, ultimately generating high-quality noise reduction signals.

[0091] The implementation process is as follows: When the system starts working, after the environmental sound collection module and motion state perception module collect data, for basic data processing tasks with high real-time requirements, such as preliminary screening and format conversion of the collected data, they are quickly calculated and processed by the high-performance mobile CPU of the mobile phone combined with GPU acceleration. For convolutional neural network and recurrent neural network operation tasks that require in-depth analysis, these complex computing tasks are transmitted to the cloud server via wireless network. The cloud server processes these tasks with its powerful computing ability and returns the results to the mobile phone via the network. The mobile phone then generates noise reduction signals based on these results.

[0092] In a specific usage scenario, when the user is on the subway and the surrounding environmental noise is complex and changing in real time. At this time, the local processor of the mobile phone quickly processes the data of the environmental sound collection module and the motion state perception module, and timely captures the instantaneous changes in noise and the motion state of the mobile phone. For complex deep learning model computing tasks, such as analyzing the characteristics of different frequency noises and predicting the trend of noise changes, they are handed over to the cloud server for processing. The cloud server quickly completes the calculation and returns the results, and the mobile phone generates accurate noise reduction signals accordingly, effectively reducing the noise in the subway and providing a quiet auditory environment for the user. In scenarios with more complex noises such as outdoor construction sites, this distributed computing architecture can also operate efficiently, ensuring that the system can excellently complete the noise reduction task in various complex environments.

[0093] Preferably, the system further includes:

[0094] A power management module that adopts a dynamic power management strategy for the power consumption characteristics of each module of the active noise cancellation system. When the ambient noise is low or the mobile phone is in a stationary state, it reduces the operation frequency of the deep learning processing module, reduces the power consumption of the GPU, and when the amount of data collected by the microphone and sensor is small, it reduces their sampling frequency to save power. At the same time, it optimizes the power output of the speaker execution module and dynamically adjusts the output power according to the noise cancellation requirements.

[0095] Specifically, in this active noise cancellation control system of the smart phone, the power management module plays an indispensable role. Its significance lies in that each module of the active noise cancellation system consumes power during operation, and the power management module adopts a dynamic power management strategy, which can effectively improve the battery life of the mobile phone and avoid excessive power consumption due to the active noise cancellation function.

[0096] In terms of the implementation process, the system continuously monitors the ambient noise situation and the state of the mobile phone. When it detects that the ambient noise is low, which means that the real-time performance and processing intensity requirements for noise cancellation are reduced, or the mobile phone is in a stationary state and the data change of the motion state perception module is small, the power management module will reduce the operation frequency of the deep learning processing module, reduce the resource consumption of the GPU for complex operations, thereby reducing power consumption. When the amount of data collected by the microphone and sensor is small, it reduces their sampling frequency, which can not only meet the basic data collection requirements but also reduce power consumption. For the speaker execution module, the power management module dynamically adjusts its power output according to the current noise cancellation requirements. If the noise cancellation requirement is low, it reduces the output power.

[0097] In a specific usage scenario, for example, when the user is in a quiet library and the ambient noise is extremely low, the power management module starts the dynamic strategy at this time, reduces the operation frequency of the deep learning processing module, and the sampling frequencies of the microphone and sensor are also reduced accordingly, and the speaker output power is reduced. While ensuring the basic noise cancellation function, it greatly saves power. The same is true when the user places the mobile phone aside and it is stationary, avoiding unnecessary power consumption and extending the battery life of the mobile phone, so that when the user needs to use the active noise cancellation function, the mobile phone still has sufficient power.

[0098] On the other hand, this embodiment also provides an active noise cancellation control method for a smart phone, which includes:

[0099] Step S1: Synchronously collect ambient sound signals, mobile phone motion data, and environmental parameter data at a set frequency through an ambient sound collection module, a motion state perception module, and other auxiliary sensors. The ambient sound collection module collects sound signals at a sampling rate of not less than 48 kHz, the motion state perception module collects motion data at a frequency of not less than 100 Hz, and other sensors collect data at their respective optimal response frequencies, and transmit these data to the mobile phone main processor through the corresponding communication interfaces;

[0100] Step S2: Preprocess the collected raw data, including removing noise interference, data calibration, and normalization operations. For ambient sound signals, use digital filtering technology to remove high-frequency or low-frequency noise; calibrate the motion data to eliminate sensor drift errors; normalize all data so that the data is within the same numerical range;

[0101] Step S3: Input the preprocessed data into the deep learning processing module, and use a hybrid model combining a convolutional neural network and a recurrent neural network to extract spatial and temporal features in the data, analyze the type, intensity, and change trend of ambient noise, and the model optimizes the model parameters by continuously learning training data;

[0102] Step S4: Generate a noise reduction signal with an amplitude equal to and a phase opposite to that of the ambient noise according to the output result of the deep learning processing module. Use an adaptive filter to update the filter coefficients using the least mean square algorithm or the recursive least squares algorithm according to the real-time change of the ambient noise, and adjust the noise reduction signal;

[0103] Step S5: Amplify the generated noise reduction signal to an appropriate power level through an audio power amplifier, drive the micro speaker of the noise reduction execution module to output, and cancel the ambient noise. At the same time, according to the active feedback control mechanism, collect the mixed signal after the speaker output in real time, and feedback it to the noise reduction signal generation step to further optimize and adjust the noise reduction signal;

[0104] Step S6: Read the personalized settings of the user in the user preference setting module, including the noise reduction intensity level and the noise reduction priority for specific frequency noises, and adjust the relevant parameters in the deep learning processing step and the noise reduction signal generation step according to the user settings to make the active noise reduction system meet the user's personalized needs;

[0105] Step S7: Use the scene recognition module to identify the environmental scene where the mobile phone is located. According to the recognition result, call the corresponding configuration from the pre-stored noise reduction parameter configuration files for different scenes, and optimize and adjust the parameters in the entire active noise reduction control process to meet the noise reduction requirements in different environmental scenes.

[0106] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to cover these changes and modifications.

Claims

1. An active noise cancellation control system for a smart phone, characterized in that, including: The system includes: Ambient sound acquisition module: Multiple high-sensitivity microphones are set at different positions of the mobile phone to form a spatial array, including three omnidirectional microphones and two directional microphones, to collect ambient sound signals omni-directionally and multi-angularly. The microphones have broadband response characteristics, with a frequency response range of 20 Hz - 20 kHz, and a sampling rate of not less than 48 kHz; Motion state perception module: Integrating an accelerometer and a gyroscope, it monitors the acceleration and angular velocity motion parameters of the mobile phone in real time. The measurement range of the accelerometer is ±16 g, and the resolution reaches 0.001 g; the measurement range of the gyroscope is ±2000 ° / s, and the resolution is 0.0625 ° / s. This module is connected to the main processor of the mobile phone through the I2C or SPI communication interface and transmits the collected motion data at a frequency of not less than 100 Hz; Deep learning processing module: Adopting a hybrid model architecture combining a convolutional neural network and a recurrent neural network, the convolutional neural network part includes a convolutional layer and a pooling layer, which are used to extract the spatial features in the ambient sound signal. Each convolutional layer is followed by a ReLU activation function; the pooling layer uses the maximum pooling method, and the pooling kernel size is 2x2. The recurrent neural network part uses long short-term memory network units with 256 memory units, which are used to process the time series features of the sound signal and capture the dynamic change law of noise. This module uses the mobile phone GPU to accelerate the calculation and performs real-time processing and analysis on the collected sound and motion data without affecting the normal use performance of the mobile phone.

2. The active noise cancellation control system of a smart phone according to claim 1, characterized in that The system also includes a noise reduction signal generation module: Based on the output result of the deep learning processing module, it generates a noise reduction signal with the same amplitude and opposite phase as the ambient noise, and adjusts the noise reduction signal through an adaptive filter. This filter uses the least mean square algorithm or the recursive least squares algorithm to update the filter coefficients in real time according to the change of the ambient noise. At the same time, combined with the audio power amplifier of the mobile phone, the noise reduction signal is amplified to an appropriate power level to drive the speaker to output; Noise reduction execution module: Composed of multiple micro speakers, distributed near the ear or the sound-emitting area of the mobile phone. The speakers adopt a flat design to adapt to the limited space inside the mobile phone. The frequency response range matches that of the microphones in the ambient sound acquisition module, which is 20 Hz - 20 kHz, and automatically adjusts the sound emission mode and volume of the speakers according to different usage scenarios of the mobile phone; The ambient sound acquisition module also includes: A barometric pressure sensor, which is used to detect the change of ambient barometric pressure and assist in judging the type of ambient noise. The measurement range of this barometric pressure sensor is 300 hPa - 1100 hPa, the resolution is 0.1 hPa, and the measurement accuracy is ±0.3 hPa; A temperature sensor, which monitors the ambient temperature. The measurement range of the temperature sensor is -40 °C - 125 °C, and the accuracy is ±0.5 °C. It is connected to the main processor through the I2C interface and transmits the collected barometric pressure and temperature data to the deep learning processing module together with the sound signal for comprehensive analysis.

3. The active noise reduction control system of a smart phone according to claim 1, characterized in that, The deep learning processing module also has: Online update mechanism, using incremental learning algorithm. When the mobile phone collects new environmental sound and motion data, instead of retraining the entire model, it performs incremental updates based on the existing model parameters. Model fusion technology, which fuses multiple deep learning models with different structures or trained on different datasets. According to the noise reduction performance of each model in different scenarios, different weights are assigned to each model, and the output results of multiple models are integrated.

4. The active noise reduction control system of a smart phone according to claim 2, characterized in that, The noise reduction signal generation module further includes: Sub-band decomposition unit, which divides the collected environmental sound signal and the generated noise reduction signal into multiple sub-bands according to frequency. The frequency range of 0Hz - 20kHz is divided into 16 sub-bands, and noise reduction processing is performed separately within each sub-band. According to the noise characteristics of different sub-bands, the adaptive filter coefficients are independently adjusted. Phase compensation circuit, which is used to precisely compensate the phase of the noise reduction signal for accurate alignment of the noise reduction signal and environmental noise in the time domain. The phase compensation range is ±180°, and the resolution is 0.1°.

5. The active noise reduction control system of a smart phone according to claim 2, characterized in that, The noise reduction execution module also has: Active feedback control mechanism, which sets an additional microphone near the speaker to collect the mixed signal of the noise reduction signal output by the speaker and the residual environmental noise in real time. This mixed signal is fed back to the noise reduction signal generation module for comparison and analysis with the original environmental sound signal, and further adjusts the amplitude and phase of the noise reduction signal to form a closed-loop control. Multi-mode sound generation strategy, which switches different sound generation modes according to the usage scenario of the mobile phone and user needs. In the call scenario, a mono sound generation mode is adopted to focus on canceling the noise in the call direction. In the music playback scenario, it switches to the stereo sound generation mode to ensure the stereo effect of the music while reducing noise. In a noisy outdoor environment, the speaker volume is automatically increased, and the low-frequency noise reduction effect is enhanced.

6. The active noise cancellation control system of a smart phone according to claim 1, characterized in that, The system also includes: User preference setting module, where the user can perform personalized settings on the active noise reduction system through the mobile phone application. The user selects different noise reduction intensity levels, such as mild, moderate, and severe noise reduction, and customizes the noise reduction priority for certain specific frequency noises. The system will adjust the parameters of the deep learning processing module and the noise reduction signal generation module according to the user settings to meet the user's personalized needs. Scene recognition module, which uses machine learning algorithms to identify the environmental scene where the mobile phone is located based on the data collected by the environmental sound collection module, the motion state perception module, and other sensors. For different scenarios, different noise reduction parameter configuration files are pre-stored, and the system automatically calls the corresponding configuration files.

7. The active noise reduction control system of a smart phone according to claim 1, characterized in that, The system adopts a distributed computing architecture: Some computing tasks are undertaken by the local processor of the mobile phone. The local processor uses a high-performance mobile CPU and combines GPU acceleration to quickly execute tasks with high real-time requirements. Complex deep learning model computing tasks, such as the operations of convolutional neural networks and recurrent neural networks, are transmitted to the cloud server through the wireless network for processing, and the processing results are returned to the mobile phone for generating noise reduction signals.

8. The active noise reduction control system of a smart phone according to claim 1, characterized in that, It also includes: The power management module adopts a dynamic power management strategy for the power consumption characteristics of each module in the active noise cancellation system. When the ambient noise is low or the mobile phone is in a stationary state, it reduces the operation frequency of the deep learning processing module, reduces the power consumption of the GPU. When the amount of data collected by the microphone and sensors is small, it reduces the sampling frequency to save power. At the same time, it optimizes the power output of the speaker execution module and dynamically adjusts the output power according to the noise cancellation requirements.

9. An active noise reduction control method for a smart phone, characterized in that, The method includes: Step S1: Synchronously collect ambient sound signals, mobile phone motion data, and environmental parameter data at a set frequency through the ambient sound collection module, motion state perception module, and other auxiliary sensors. The ambient sound collection module collects sound signals at a sampling rate of not less than 48 kHz, the motion state perception module collects motion data at a frequency of not less than 100 Hz, and other sensors collect data at their respective optimal response frequencies, and transmit these data to the mobile phone main processor through the corresponding communication interfaces; Step S2: Preprocess the collected raw data, including removing noise interference, data calibration, and normalization operations. For ambient sound signals, digital filtering technology is used to remove high-frequency or low-frequency noise; calibrate the motion data to eliminate sensor drift errors; normalize all data so that the data is within the same numerical range; Step S3: Input the preprocessed data into the deep learning processing module, and use a hybrid model combining convolutional neural network and recurrent neural network to extract spatial and temporal features in the data, analyze the type, intensity, and change trend of ambient noise, and the model optimizes the model parameters by continuously learning training data; Step S4: Generate a noise cancellation signal with the same amplitude and opposite phase as the ambient noise according to the output result of the deep learning processing module. Use an adaptive filter to update the filter coefficients using the least mean square algorithm or recursive least squares algorithm according to the real-time change of the ambient noise, and adjust the noise cancellation signal; Step S5: Amplify the generated noise cancellation signal to an appropriate power level through an audio power amplifier, drive the micro speaker of the noise cancellation execution module to output, cancel the ambient noise. At the same time, according to the active feedback control mechanism, collect the mixed signal after the speaker output in real time, and feedback it to the noise cancellation signal generation step to further optimize and adjust the noise cancellation signal.

10. The active noise reduction control method of a smart phone according to claim 9, characterized in that, It also includes: Step S6: Read the personalized settings of the user in the user preference setting module, including the noise cancellation intensity level and the noise cancellation priority for specific frequencies. According to the user settings, adjust the relevant parameters in the deep learning processing step and the noise cancellation signal generation step to make the active noise cancellation system meet the user's personalized needs; Step S7: Use the scene recognition module to identify the environmental scene where the mobile phone is located. According to the recognition result, call the corresponding configuration from the pre-stored noise cancellation parameter configuration files for different scenes, and optimize and adjust the parameters in the entire active noise cancellation control process to meet the noise cancellation requirements in different environmental scenes.

Images