A method, system, device and medium for data loss compensation of a digital audio conference system

Abstract

The application provides a data loss compensation method, system, device and medium of a digital audio conference system, comprising: acquiring real-time audio data stream in the digital audio conference system, inputting the real-time audio data stream into a dynamic cache area through a dynamic cache technology to obtain dynamic cache area data; detecting a loss type of a sampling point time sequence interval of the dynamic cache area data to obtain a loss type flag, wherein the loss type flag comprises a single-point loss flag and / or a whole-frame loss flag; determining a compensation strategy according to the loss type flag, and compensating the digital audio signal based on the compensation strategy to obtain compensation data, wherein the compensation data comprises a single-point compensation value and / or frame compensation data; and mapping the compensation data to the original audio time sequence position in the real-time audio data stream for interpolation to obtain a continuous audio output stream. The method can maintain the phase continuity of the audio waveform and improve the call quality and auditory experience in a complex transmission scenario.

Description

Technical Field

[0001] This invention belongs to the field of digital audio technology, and in particular relates to a method, system, device and medium for data loss compensation in a digital audio conferencing system. Background Technology

[0002] With the development of audio and video technologies, digital audio systems, as communication systems for real-time data transmission, have been widely used, especially high-quality audio systems, which have high requirements for latency and noise interference. In this field, existing technologies address the problem of missing data by employing methods such as adding filters, fixed compensation, and storing audio signals and compensating based on their characteristics. Adding filters aims to reduce abnormal frequencies, fixed compensation uses preset methods to compensate, while storing audio signals and compensating based on their characteristics relies on sound features to perform the compensation operation.

[0003] However, these methods have significant limitations: interpolation algorithms introduce harmonic distortion at abrupt changes in speech (such as plosives), disrupting the continuity of sound quality; retransmission mechanisms inevitably introduce high latency due to network interaction, causing breaks in the rhythm of voice conversations; and simple compensation by directly copying historical data generates audible high-frequency transient noise (such as "crackling") at signal transition points, severely impairing the naturalness of the sound. Especially in low-bandwidth, high-interference scenarios such as wireless transmission, the accuracy deficiencies and latency issues of existing technologies are further amplified, making it difficult to simultaneously meet the dual requirements of real-time performance and fidelity, thus affecting the reliability and user experience of digital audio conferencing systems. Summary of the Invention

[0004] Based on this, it is necessary to provide a data loss compensation method, system, device and medium for a digital audio conferencing system to address the above-mentioned technical problems. This method can maintain the phase continuity of the audio waveform while eliminating high-frequency transient noise, and achieve zero additional latency by relying on high-speed pipeline processing, thereby improving call quality and listening experience in complex transmission scenarios.

[0005] Firstly, this application provides a method for compensating for data loss in a digital audio conferencing system, including:

[0006] The system acquires real-time audio data streams from digital audio conferencing systems and inputs these streams into a dynamic buffer using dynamic caching technology to obtain data from the dynamic buffer.

[0007] Missing type detection is performed on the sampling point time interval of the dynamic buffer data to obtain missing type flags, which include single-point missing flags and / or whole-frame missing flags.

[0008] The compensation strategy is determined based on the missing type flag, and the digital audio signal is compensated based on the compensation strategy to obtain compensation data, which includes single-point compensation values ​​and / or frame compensation data.

[0009] The compensation data is mapped to the original audio timing position in the real-time audio data stream for interpolation to obtain a continuous audio output stream.

[0010] In one embodiment, a missing type detection is performed on the sampling point time interval of the dynamic buffer data to obtain a missing type flag, including:

[0011] The actual receiving interval is calculated by clocking adjacent sampling points in the dynamic buffer data, where the actual receiving interval is the arrival time difference between the current sampling point and the previous sampling point.

[0012] When the actual reception interval exceeds the preset standard sampling period, a single point of loss determination is performed to obtain a single point of loss flag.

[0013] The single-point missing flags obtained continuously are counted and accumulated by a counter. When the count value reaches the preset frame length, the frame missing flag is converted to obtain the whole frame missing flag and the counter is reset.

[0014] In one embodiment, a compensation strategy is determined based on a missing type flag, and the digital audio signal is compensated based on the compensation strategy to obtain compensation data, including:

[0015] Based on the single-point missing flag, the Nth valid sampling point before the missing point and its amplitude and phase characteristics are extracted from the dynamic buffer data to obtain the waveform reconstruction reference parameters.

[0016] The pre-stored sine wave signal template is called, and phase synchronization processing is performed based on the waveform reconstruction reference parameters to obtain a phase-aligned sine wave template;

[0017] Adjust the amplitude of the phase-aligned sine wave template according to the amplitude change trend of the waveform reconstruction reference parameters to obtain the single-point compensation value.

[0018] In one embodiment, a pre-stored sine wave signal template is invoked, and phase synchronization processing is performed based on waveform reconstruction reference parameters to obtain a phase-aligned sine wave template, including:

[0019] The instantaneous phase angle is calculated for the Nth valid sampling point before the missing point to obtain the current phase reference;

[0020] The initial phase of the pre-stored sine wave signal template is offset and compensated to generate a synchronous sine wave that is aligned with the current phase reference.

[0021] The phase adjustment amount in the offset compensation process is equal to that of the current phase reference.

[0022] In one embodiment, a compensation strategy is determined based on a missing type flag, and the digital audio signal is compensated based on the compensation strategy to obtain compensated data. The method further includes:

[0023] Based on the missing frame flag, the most recent complete historical frame is copied from the dynamic buffer data to obtain the frame data to be processed;

[0024] Low-pass filtering is performed on the start and end data segments of the frame data to be processed to obtain a smooth frame transition region.

[0025] The frequency band smoothing frame transition area is covered to the missing frame position indicated by the whole frame missing flag to replace the data and obtain frame compensation data.

[0026] In one embodiment, a real-time audio data stream from a digital audio conferencing system is acquired, and the real-time audio data stream is input into a dynamic buffer using dynamic caching technology to obtain dynamic buffer data, including:

[0027] A circular buffer queue is used to store the real-time audio input stream using a sliding window, generating a continuously updated historical data set.

[0028] When new data is written, the oldest data that has exceeded the preset time limit is overwritten to generate a dynamic cache with a constant capacity.

[0029] Establish an index mapping table for the historical data set to access the Kth sampling point before the missing point and the most recent complete historical frame.

[0030] In one embodiment, the compensation data is mapped to the original audio timing position in the real-time audio data stream for interpolation to obtain a continuous audio output stream, including:

[0031] Obtain the raw audio timing table of the real-time audio data stream. The raw audio timing table is used to record the timing position information of each sampling point in the real-time audio data stream. The timing position information includes the timestamp of the sampling point and its sequence number in the data stream.

[0032] Based on the missing type flag, the target temporal position corresponding to the missing data in the real-time audio data stream is determined. If the missing type flag is a single-point missing flag, the target temporal position is the temporal position of a single sampling point; if the missing type flag is an entire frame missing flag, the target temporal position is the continuous temporal position interval corresponding to a frame of data.

[0033] The compensation data is inserted into the real-time audio data stream according to the target time sequence position, and the inserted data stream is interpolated to obtain a preliminary fused audio stream;

[0034] The initial fused audio stream is transmitted to a preset low-pass filter. The low-pass filter filters out abnormal high-frequency signals generated by data interpolation in the initial fused audio stream to obtain a continuous audio output stream. The abnormal high-frequency signals are signals that exceed the normal audio frequency range when the compensation data is connected to the original valid data.

[0035] Secondly, this application also provides a data loss compensation system for a digital audio conferencing system, comprising:

[0036] The audio data input module is used to acquire real-time audio data streams from digital audio conferencing systems. It inputs the real-time audio data streams into a dynamic buffer area using dynamic caching technology to obtain dynamic buffer area data.

[0037] The data missing detection module is used to perform missing type detection on the sampling point time interval of the dynamic buffer data and obtain the missing type flag, which includes a single point missing flag and / or an entire frame missing flag.

[0038] The data classification and compensation module is used to determine the compensation strategy based on the missing type flag, and to compensate the digital audio signal based on the compensation strategy to obtain compensation data, which includes single-point compensation values ​​and / or frame compensation data.

[0039] The data output module is used to map the compensation data to the original audio timing position in the real-time audio data stream for interpolation, so as to obtain a continuous audio output stream.

[0040] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the data loss compensation method of the above-described digital audio conferencing system.

[0041] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the data loss compensation method of the above-described digital audio conferencing system.

[0042] The aforementioned data loss compensation method, system, device, and medium for a digital audio conferencing system utilizes dynamic caching technology to process the input audio stream in real time, constructing a continuously updated buffer dataset to provide a latency-free processing foundation for high-speed pipelines. Based on intelligent analysis of sampling point timing intervals, it accurately distinguishes between single-point and whole-frame loss types, laying the groundwork for differentiated compensation decisions. For each identified loss type, it dynamically invokes an appropriate compensation strategy to obtain single-point compensation values ​​or frame compensation data, ensuring waveform continuity and frequency band stability at the algorithmic level. The compensation data is mapped to the precise timing position of the original audio stream for interpolation output, eliminating phase jumps caused by clock deviations. This technical solution can simultaneously suppress high-frequency transient noise and maintain waveform phase integrity at the signal level through a typified compensation mechanism. Simultaneously, at the system level, relying on the hardware-friendly design of dynamic caching and timing interpolation, it improves the high-speed processing efficiency across the entire link, enhancing the naturalness of calls and the auditory experience in complex transmission scenarios. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 A flowchart illustrating a data loss compensation method for a digital audio conferencing system provided in an embodiment of the present invention;

[0045] Figure 2 A flowchart illustrating a method for detecting missing digital audio data according to an embodiment of the present invention;

[0046] Figure 3 This is a schematic diagram of the structure of a data loss compensation system for a digital audio conferencing system provided in an embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] In one embodiment, such as Figure 1As shown, a data loss compensation method for a digital audio conferencing system is provided. This embodiment illustrates the method by applying it to a terminal. It is understood that the system can also be applied to a server, and to an operating system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps 101 to 104:

[0049] Step 101: Obtain the real-time audio data stream from the digital audio conferencing system, and input the real-time audio data stream into the dynamic buffer area using dynamic caching technology to obtain dynamic buffer area data.

[0050] Specifically, the method first acquires real-time generated digital audio data streams from the audio acquisition end of the digital audio conferencing system (such as the microphone array of the conferencing terminal, external audio input devices, etc.). These data streams are digital signals generated after the audio acquisition device samples, quantizes, and encodes analog audio signals such as human voices and ambient sounds in the conferencing scene. They carry key information such as the amplitude, phase, and timing of the conferencing audio. To avoid data loss due to instantaneous data volume fluctuations or processing delays during real-time data transmission, this method employs dynamic caching technology to store the acquired real-time audio data stream. The dynamic buffer can be implemented based on a First-In-First-Out (FIFO) storage structure or a circular buffer queue structure. For example, using a FIFO storage structure, the method writes data sequentially to the dynamic buffer according to the generation sequence of the real-time audio data stream. Simultaneously, when the buffer reaches its preset capacity limit, it automatically overwrites the oldest written historical data, ensuring that the buffer always stores the latest audio data. If a circular buffer queue structure is used, the method dynamically manages data writing and reading through the head and tail pointers of the queue. New data is written from the tail of the queue, and reading operations are performed from the head. When the tail pointer reaches the queue's storage boundary, it automatically jumps to the beginning of the queue, achieving circular data storage. Through the application of this dynamic caching technology, the method stably inputs the real-time audio data stream into the dynamic buffer. The resulting dynamic buffer data can completely preserve the temporal relationship and signal characteristics of recent real-time audio data, providing accurate and continuous data source support for subsequent data loss detection and effectively avoiding detection errors caused by data transmission interruptions or processing delays.

[0051] Step 102: Perform missing type detection on the sampling point time interval of the dynamic buffer data to obtain missing type flags, wherein the missing type flags include single-point missing flags and / or whole-frame missing flags.

[0052] For example, the method first extracts the stored dynamic buffer data from the dynamic buffer. This data exists in the form of discrete sampling points, each of which carries corresponding timing information (such as a generation timestamp or sampling sequence number). Based on the periodic sampling characteristics of digital audio signals, the method first determines the system's preset standard sampling period (this period is determined by the sampling rate of the digital audio conferencing system; for example, when the system uses a 48kHz sampling rate, the standard sampling period is 1 / 48000 seconds). Then, it calculates the actual timing interval between two adjacent sampling points in the dynamic buffer data one by one. The calculation method is to obtain the timing information between the current sampling point and the previous sampling point (if it is a timestamp, the timestamp difference is calculated; if it is a sampling sequence number, it is converted into an actual time interval according to the sampling rate) to obtain the actual timing interval between adjacent sampling points. Furthermore, this method compares the calculated actual timing interval with a preset standard sampling period. When the actual timing interval of a certain adjacent sampling point is greater than the standard sampling period but less than twice the standard sampling period, it is determined that a single point of data is missing. At this time, the method generates a single-point missing flag and marks the corresponding missing sampling point location. When the actual timing intervals of multiple consecutive adjacent sampling points all meet the single-point missing judgment condition, and the number of consecutively missing sampling points reaches the preset data length of one frame of the digital audio conferencing system (the data length of one frame is determined by the system protocol or audio encoding format, for example, each frame contains 256 sampling points), the method determines that an entire frame of data is missing. At this time, an entire frame missing flag is generated and the corresponding missing frame interval is marked. Through the above detection method based on the timing interval of sampling points, this method can accurately identify the type and location of data missing. The obtained single-point missing flag and / or entire frame missing flag can directly provide a clear basis for determining the subsequent compensation strategy, effectively avoiding the misjudgment problem caused by relying solely on changes in data amplitude in traditional detection methods.

[0053] Step 103: Determine the compensation strategy based on the missing type flag, and compensate the digital audio signal based on the compensation strategy to obtain compensation data, which includes single-point compensation values ​​and / or frame compensation data.

[0054] Specifically, when the missing type flag detected by this method is a single-point missing flag, the compensation strategy determined is a single-point waveform reconstruction strategy based on the continuity of the audio signal. Specifically, this method first extracts multiple consecutive valid sampling points before the missing sampling point from the dynamic buffer data. These sampling points carry the amplitude change trend and phase characteristics of the audio signal before the missing point. This method constructs a waveform prediction model of the audio signal at the missing point by calculating and analyzing the amplitude and phase of these valid sampling points. At the same time, it calls the standard sine wave signal template pre-stored by the system (this template is selected based on the commonly used audio frequency range of digital audio conferencing systems and has complete periodic waveform characteristics). The constructed waveform prediction model is phase aligned and amplitude matched with the standard sine wave signal template to generate a single-point compensation value consistent with the characteristics of the audio signal before and after the missing point. When the detected missing type flag is a whole-frame missing flag, the compensation strategy determined by this method is a frame data filling strategy based on historical frame reuse. This method extracts the most recent complete audio data frame before the missing frame from the dynamic buffer data. Since this frame is temporally adjacent to the missing frame, its audio signal frequency, amplitude, and phase characteristics should have strong continuity with the missing frame. This method performs a slight amplitude smoothing process on the extracted complete historical frame data (to avoid signal abrupt changes caused by direct reuse), and uses the processed historical frame data as frame compensation data. Through the above targeted compensation strategy, the single-point compensation value and / or frame compensation data generated by this method can fit the characteristics of the original audio signal to the greatest extent, effectively avoiding the problems of timbre distortion or phase breakage in the compensated audio signal.

[0055] Step 104: Map the compensation data to the original audio timing position in the real-time audio data stream and interpolate to obtain a continuous audio output stream.

[0056] Specifically, the method obtains the original timing planning information of the real-time audio data stream from the timing management module of the digital audio conferencing system. This information includes the preset timing position (such as timestamp or sampling number) of each sampling point in the data stream. Then, based on the previously obtained missing type flag, the method determines the target timing position corresponding to a single-point missing or an entire frame missing in the original timing planning information. A single-point missing corresponds to the timing position of a single sampling point, while an entire frame missing corresponds to a frame timing interval composed of multiple consecutive sampling points. Further, the method maps the generated compensation data (single-point compensation value or frame compensation data) to the missing positions in the real-time audio data stream according to the target timing positions. To ensure a smooth transition between the compensation data and the normal audio data before and after the missing position, the method uses linear interpolation or nonlinear interpolation algorithms (such as interpolation algorithms based on the derivative of the audio signal) to interpolate the mapped data stream, enabling a seamless transition in amplitude and phase between the compensation data and the normal data before and after it. Ultimately, through the aforementioned mapping and interpolation processes, this method yields a continuous audio output stream without data loss or signal abrupt changes. This output stream meets the high audio quality requirements of digital audio conferencing systems, ensuring that the audio signals received by conference participants are clear and smooth.

[0057] The aforementioned data loss compensation method for a digital audio conferencing system utilizes dynamic caching technology to process the input audio stream in real time, constructing a continuously updated buffer dataset to provide a latency-free processing foundation for high-speed pipelines. Based on intelligent analysis of sampling point timing intervals, it accurately distinguishes between single-point and whole-frame loss types, laying the groundwork for differentiated compensation decisions. For each identified loss type, it dynamically invokes an appropriate compensation strategy to obtain single-point compensation values ​​or frame compensation data, ensuring waveform continuity and frequency band stability at the algorithmic level. The compensation data is mapped to the precise timing position of the original audio stream for interpolation output, eliminating phase jumps caused by clock deviations. This technical solution can simultaneously suppress high-frequency transient noise and maintain waveform phase integrity at the signal level through a typified compensation mechanism. Simultaneously, at the system level, relying on the hardware-friendly design of dynamic caching and timing interpolation, it improves the high-speed processing efficiency across the entire link, enhancing the naturalness of calls and the auditory experience in complex transmission scenarios.

[0058] like Figure 2 As shown, in one embodiment, a missing type detection is performed on the sampling point time interval of the dynamic buffer data to obtain a missing type flag, including:

[0059] Step 201: Calculate the actual receiving interval by performing clock interval calculation based on the receiving timestamps of adjacent sampling points in the dynamic buffer data. The actual receiving interval is the arrival time difference between the current sampling point and the previous sampling point.

[0060] Specifically, the method extracts stored dynamic buffer data from the dynamic buffer. This data is obtained after storing the real-time audio data stream of the digital audio conferencing system using dynamic caching technology. Each sampling point carries a unique reception timestamp, which is generated in real time by the clock module of the digital audio conferencing system when the sampling point is acquired and transmitted to the dynamic buffer. This timestamp accurately reflects the specific time when each sampling point arrives at the dynamic buffer. To obtain the actual reception interval between adjacent sampling points, the method selects two adjacent sampling points sequentially according to the storage order of the dynamic buffer data: the current sampling point and the previous sampling point. Then, it calls the time difference calculation logic to subtract the reception timestamp of the previous sampling point from the reception timestamp of the current sampling point. This subtraction operation yields the time difference between the two sampling points, which is the actual reception interval between adjacent sampling points. For example, this method can extract the received timestamp and calculate the difference through hardware-level timing logic circuits, or process the received timestamp data through software-level time difference algorithms. Both implementation methods can ensure the accuracy of the actual receiving interval calculation, providing reliable time dimension data support for subsequent missing type determination and avoiding misjudgment of missing detection due to time data errors.

[0061] Step 202: When the actual receiving interval exceeds the preset standard sampling period, a single point missing determination is performed to obtain a single point missing flag.

[0062] Specifically, this method first obtains the pre-set standard sampling period of the digital audio conferencing system. This standard sampling period is determined by the system's audio sampling rate and represents the theoretical time interval between two adjacent sampling points under normal operating conditions. Its value is determined during system initialization based on the conference audio quality requirements and stored in a designated storage area. The method then compares the previously calculated actual reception interval between adjacent sampling points with the preset standard sampling period. If the actual reception interval is greater than the standard sampling period, it indicates that the time interval between the current adjacent sampling points exceeds the expected range under normal system operation. Combined with the continuous transmission characteristics of digital audio signals, it can be determined that a single sampling point is missing between the current adjacent sampling points. At this point, the method triggers a single-point-of-missing-point (SPO) determination logic, generating a SPO flag to identify the SPO situation. This flag can be implemented using binary code or specific data identifiers, such as using binary "1" to indicate the presence of a SPO and binary "0" to indicate no SPO; or it can use specific character identifiers, such as "SP_MISS" as the SPO flag. Through the above judgment process, this method can quickly and accurately identify single-point missing data in the dynamic cache area. The generated single-point missing flag can be directly used for further judgment of the missing type or triggering of compensation strategies, effectively improving the timeliness and accuracy of data missing detection.

[0063] Step 203: The continuously obtained single-point missing flags are counted and accumulated using a counter. When the count value reaches the preset frame length, a frame missing conversion is performed to obtain a whole frame missing flag and the counter is reset.

[0064] Specifically, a counter is first initialized, with its initial count value set to zero. This counter possesses functions for counting accumulation, value judgment, and resetting. It can be implemented through a hardware counter circuit or a software-level counting variable. For example, an FPGA-based hardware counter circuit can achieve high-speed counting accumulation operations, adapting to the real-time data processing needs of digital audio conferencing systems. Alternatively, an integer variable in software programming can be used as a counter, implementing the counting function through a loop accumulation statement. After generating a single-point missing flag, the method determines whether the single-point missing flag is consecutive. If the previous sampling point's missing result is a single-point missing flag and the current sampling point's missing result is also a single-point missing flag, then it is determined that consecutive single-point missing flags have been obtained. In this case, the method controls the counter to accumulate the consecutive single-point missing flags; that is, each time consecutive single-point missing flags are detected, the counter's count value increases by 1. Simultaneously, the method obtains the preset frame length of the digital audio conferencing system. This frame length is the number of sampling points contained in one frame of audio data as defined by the system, a fixed value determined based on parameters such as the system's audio encoding format and transmission bandwidth. This method compares the current counter value with the preset frame length in real time. If the counter value reaches the preset frame length, it indicates that the number of consecutive missing sampling points has reached the number of sampling points for one frame of audio data. At this point, it can be determined that there is a missing frame of data in the dynamic buffer. The method then triggers the frame missing conversion logic, converting the judgment result of consecutive single-point missing points into the judgment result of whole-frame missing points, and generating a whole-frame missing flag to identify the whole-frame missing situation. The implementation of this flag can refer to the single-point missing flag, for example, it can use binary "11" to represent the existence of whole-frame missing points, or use the character identifier "FR_MISS". After generating the whole-frame missing flag, the method controls the counter to perform a reset operation, restoring the counter value to the initial zero value, so that the counter can restart the counting and accumulation of subsequent consecutive single-point missing flags. Through the aforementioned counter accumulation and frame missing conversion process, this method can accurately identify whole-frame missing situations from continuous single-point missing situations, avoiding misjudging whole-frame missing situations as multiple independent single-point missing situations. The generated whole-frame missing flag provides an accurate basis for triggering subsequent targeted frame compensation strategies. At the same time, the counter reset operation ensures the accuracy and independence of subsequent missing situation detection.

[0065] In one embodiment, a compensation strategy is determined based on a missing type flag, and the digital audio signal is compensated based on the compensation strategy to obtain compensation data, including:

[0066] Based on the single-point missing flag, the Nth valid sampling point before the missing point and its amplitude and phase characteristics are extracted from the dynamic buffer data to obtain the waveform reconstruction reference parameters.

[0067] Specifically, after detecting a single missing point, this method first determines the specific location of the missing point in the dynamic buffer data. This location is precisely located using the sampling point timing information (such as the receiving timestamp or sampling sequence number) recorded during the previous missing point type detection process, ensuring accurate locking of the timing coordinates corresponding to the missing point. Subsequently, based on these timing coordinates, the method filters out valid sampling points preceding the missing point from the dynamic buffer data. During the filtering, continuous sampling points without data anomalies (such as amplitude abrupt changes or phase breaks) are prioritized. Finally, the method determines the Nth valid sampling point before the missing point. The value of N is determined based on the continuity characteristics of the digital audio signal and the system's processing accuracy requirements. The selected N valid sampling points can completely reflect the variation pattern of the audio signal before the missing point. To obtain the amplitude and phase characteristics of these valid sampling points, the method analyzes the digital signal value of each valid sampling point. The amplitude characteristic is obtained by calculating the absolute value or square of the digital signal value of the sampling point, directly reflecting the intensity change of the audio signal; the phase characteristic is calculated by performing a Fourier transform or Hilbert transform on the digital signal value of the sampling point, reflecting the vibration phase state of the audio signal. This method integrates the amplitude and phase data of the extracted Nth valid sampling point, along with the correlation between these data and the sampling time sequence, to form waveform reconstruction reference parameters for subsequent waveform reconstruction. These parameters can be stored, for example, by constructing amplitude and phase sequences, ensuring that the parameters completely preserve the key features of the audio signal before the missing point. Through the above extraction and calculation process, the waveform reconstruction reference parameters obtained by this method provide an accurate original signal reference for the generation of subsequent compensation data, avoiding the decoupling of the compensation signal from the original signal due to inaccurate reference data.

[0068] The pre-stored sine wave signal template is invoked, and phase synchronization processing is performed based on the waveform reconstruction reference parameters to obtain a phase-aligned sine wave template.

[0069] Specifically, this method first retrieves a pre-stored sine wave signal template from a fixed storage area (such as a ROM storage unit) of the digital audio conferencing system. This template is a complete periodic sine wave data pre-generated based on the commonly used audio frequency range of the digital audio conferencing system (covering typical human voice frequencies and ambient sound frequencies in a conference scenario). It contains multiple complete periodic sine wave sampling points, and each sampling point carries corresponding phase information, serving as a standard reference carrier for audio signal waveform reconstruction. After retrieving the sine wave signal template, the method uses the phase characteristics in the waveform reconstruction reference parameters as the core reference to initiate the phase synchronization processing flow: extracting the phase value of the last valid sampling point before the missing point in the waveform reconstruction reference parameters, and using this phase value as the target reference phase for phase synchronization; adjusting the phase offset of the pre-stored sine wave signal template, and calculating the difference between the phase of each sampling point in the sine wave signal template and the target reference phase through a phase compensation algorithm during the adjustment process, and then correcting the phase of the sine wave signal template point by point based on this difference, so that the phase of the corresponding sampling point in the corrected sine wave signal template is consistent with the target reference phase. For example, this method can employ a linear phase offset algorithm to achieve phase synchronization. By adding a fixed phase offset to the phase function of the sinusoidal signal template, the overall template phase is directly adjusted to align with the target reference phase. Alternatively, an adaptive phase matching algorithm can be used. Based on the phase change trends of multiple effective sampling points in the waveform reconstruction reference parameters, the phase of the sinusoidal signal template is dynamically adjusted to ensure that the template phase not only matches the target reference phase but also matches the phase change pattern before the missing point. Through the above phase synchronization processing, this method obtains a phase-aligned sinusoidal template whose phase characteristics perfectly match the phase characteristics of the audio signal before the missing point, laying the foundation for subsequent generation of phase-continuous compensation data.

[0070] Adjust the amplitude of the phase-aligned sine wave template according to the amplitude change trend of the waveform reconstruction reference parameters to obtain the single-point compensation value.

[0071] Specifically, this method first analyzes the amplitude characteristics in the waveform reconstruction reference parameters, extracting the variation law of the amplitude data of the Nth effective sampling point before the missing point with the sampling time sequence, i.e., the amplitude change trend. Specifically, this method obtains a fitting curve of amplitude change over time by performing linear or nonlinear fitting (such as a least squares-based fitting algorithm) on the amplitude data of the N effective sampling points. This curve can intuitively reflect whether the amplitude of the audio signal before the missing point shows an increasing, decreasing, or stable trend. After obtaining the amplitude change trend, this method uses a phase-aligned sine wave template as a basis and dynamically adjusts the amplitude of the template according to the fitted amplitude change trend: if the amplitude change trend is increasing, the amplitude of the phase-aligned sine wave template is gradually increased according to the increasing rate; if the amplitude change trend is decreasing, the amplitude of the template is gradually decreased according to the decreasing rate; if the amplitude change trend is stable, the template amplitude is kept consistent with the amplitude of the last effective sampling point before the missing point. During the adjustment process, this method calculates and corrects the amplitude of each sampling point of the phase-aligned sine wave template using an amplitude scaling algorithm, ensuring that the corrected template amplitude perfectly matches the amplitude variation pattern before the missing point. Subsequently, based on the temporal position of the single-point missing, the method extracts the amplitude of a single sampling point corresponding to the timing of the missing point from the amplitude-adjusted sine wave template. This amplitude is the single-point compensation value used to fill the single-point missing. For example, this method can use a proportional scaling algorithm to scale the amplitude of the phase-aligned sine wave template according to a proportional coefficient based on the amplitude variation trend, directly obtaining the appropriate amplitude; alternatively, it can use an interpolation calculation algorithm to interpolate and optimize the amplitude of the adjusted template based on the amplitude and variation trend of the effective sampling points before and after the missing point, further improving the matching degree between the single-point compensation value and the original signal. Through the above amplitude adjustment and sampling point extraction, the single-point compensation value obtained by this method not only maintains continuity with the original audio signal in phase but also completely follows the variation pattern of the original signal in amplitude, accurately filling the single-point missing and avoiding noise caused by abrupt amplitude changes in the compensated audio.

[0072] In one embodiment, a pre-stored sine wave signal template is invoked, and phase synchronization processing is performed based on waveform reconstruction reference parameters to obtain a phase-aligned sine wave template, including:

[0073] The instantaneous phase angle is calculated for the Nth valid sampling point before the missing point to obtain the current phase reference.

[0074] Specifically, from the acquired waveform reconstruction reference parameters, the digital signal value of the Nth valid sampling point before the missing point is extracted. These valid sampling points are data previously selected from the dynamic buffer data that can reflect the continuous change pattern of the audio signal before the missing point, and their digital signal values ​​directly carry the amplitude and phase correlation information of the audio signal. To accurately calculate the instantaneous phase angle of this valid sampling point, this method uses a phase extraction algorithm based on Hilbert transform. By performing a Hilbert transform on the digital signal value of the valid sampling point, an analytical signal orthogonal to the original signal is obtained, and then the instantaneous phase angle is calculated based on the analytical signal. The instantaneous phase angle of the digital signal value of the Nth valid sampling point before the missing point is calculated using the following formula to obtain the current phase reference:

[0075]

[0076] in, Let x(n) be the instantaneous phase angle of the Nth valid sampling point before the missing point (i.e., the current phase reference), x(n) be the digital signal value of the Nth valid sampling point before the missing point, and y(n) be the orthogonal analytic signal value obtained after performing a Hilbert transform on x(n). Through the above formula, the instantaneous phase state of the audio signal before the missing point can be accurately extracted. For example, if the audio signal is in the phase increment stage, this formula can accurately capture the real-time phase change, rather than using a preset fixed initial phase, providing a unique and dynamic reference standard for the phase alignment of the subsequent sine wave template, ensuring the phase continuity between the compensation signal and the original signal.

[0077] The initial phase of the pre-stored sine wave signal template is offset and compensated to generate a synchronous sine wave that is aligned with the current phase reference.

[0078] The phase adjustment amount in the offset compensation process is equal to that of the current phase reference.

[0079] Specifically, this method first retrieves a pre-stored sine wave signal template from a fixed storage area (such as a ROM storage unit) of the digital audio conferencing system. This template is a complete periodic data pre-generated based on the typical audio frequency range in the conferencing scenario, containing multiple periods of sine wave sampling points, and each sampling point corresponds to a fixed initial phase value. Its expression can be:

[0080]

[0081] Where A is the amplitude of the sine wave, ω is the angular frequency, and t is time. The initial phase of the template, due to this initial phase Since the initial phase is a fixed value, it may deviate from the current phase reference. Therefore, this method requires offset compensation processing for the initial phase. First, the method uses the previously calculated current phase reference... Initial phase with the sine wave signal template Perform difference calculations to determine the phase deviation value; set the phase adjustment amount as follows: The initial phase of each sampling point in the sine wave signal template is corrected using a phase shift algorithm. The following formula is used to perform phase shift compensation processing on the initial phase of the pre-stored sine wave signal template to generate a synchronous sine wave:

[0082]

[0083] Where s'(t) is the synchronous sine wave generated after offset compensation. For phase adjustment amount and It can accurately compensate for the deviation between the initial phase of the sine wave template and the phase of the original signal, where A is the amplitude of the sine wave template, ω is the angular frequency of the sine wave template, and t is time. The initial phase of the sine wave signal template is used. For example, this method can achieve real-time phase adjustment of the sine wave template through a hardware-level phase offset circuit, or it can process the template data point-by-point through a software-level phase correction algorithm. Both methods ensure that the phase of the sine wave signal template is perfectly aligned with the current phase reference. Through the aforementioned offset compensation processing, the synchronous sine wave generated by this method maintains phase consistency with the audio signal before the missing point, effectively solving the problem of phase mismatch between the traditional fixed-phase template and the original signal, laying the foundation for subsequent generation of phase-continuous compensation data.

[0084] In one embodiment, a compensation strategy is determined based on a missing type flag, and the digital audio signal is compensated based on the compensation strategy to obtain compensated data. The method further includes:

[0085] Based on the missing frame flag, the most recent complete historical frame is copied from the dynamic buffer data to obtain the frame data to be processed.

[0086] Specifically, after detecting a missing frame flag, this method first uses the timing information carried by the missing frame flag (such as the start and end sampling numbers of the missing frame and the corresponding timestamp interval) to locate the historical data segment in the dynamic buffer data that is temporally adjacent to the missing frame. The dynamic buffer data is a real-time audio data stream previously stored using dynamic caching technology. Following the update logic of "new data overwrites old data," it always retains the latest few frames of complete audio data, ensuring that the complete historical frame closest in time to the missing frame can be quickly located. This method filters historical frames from the dynamic buffer data that meet the criteria of being "most recent in time" and "complete in data." The filtering criteria are that the interval between the end timing of the historical frame and the start timing of the missing frame is the smallest, and the historical frame has no abnormalities such as missing sampling points or amplitude abrupt changes, and can completely reflect the frequency, amplitude, and phase change patterns of the audio signal before the missing frame. Subsequently, the method performs a full copy of the selected most recent complete historical frame through data copying logic. During the copying process, the digital signal values ​​and timing association information of all sampling points of the historical frame are retained, ultimately obtaining the frame data to be processed for subsequent processing. For example, this method can achieve high-speed copying of historical frame data through a hardware-level DMA (Direct Memory Access) controller, reducing CPU intervention to ensure real-time performance; or it can copy historical frame data point by point through a software-level array copy algorithm. Both implementation methods can ensure that the frame data to be processed is completely consistent with the original historical frame data, providing a raw data foundation that is highly similar to the features of the missing frames for subsequent compensation processing, and avoiding obvious gaps in the compensated audio due to inaccurate historical frame data.

[0087] Low-pass filtering is performed on the start and end data segments of the frame data to be processed to obtain the frequency band smoothing frame transition area.

[0088] Specifically, the method first determines the range of the start and end data segments of the frame data to be processed. This range is determined based on the frame length of the digital audio signal and the requirements for smooth transitions. Typically, continuous sampling points from the start and end of the frame are selected to form the data segment, ensuring that this segment covers the transition area between the frame to be processed and the normal frames before and after the missing frame. Subsequently, the method analyzes the frequency characteristics of the frame data to be processed and the normal frames before and after the missing frame to identify the sources of abnormal high-frequency signals that may be generated at the transition point. Since the frame to be processed is historical data, there is a temporal gap between it and the normal frames before and after the missing frame. Direct reuse may cause abrupt changes in the amplitude or phase of the sampling points at the transition point, resulting in high-frequency transient noise that exceeds the normal audio frequency range. This type of noise needs to be suppressed through low-pass filtering. This method uses a low-pass filter adapted to the conference audio scenario. The passband frequency range of the filter covers normal human voices and common audio frequencies in conference scenarios, while the stopband frequency is set to target abnormal high-frequency signals that may be generated at the transition point, ensuring that high-frequency noise is filtered out while retaining the effective audio signal. In specific filtering processing, this method inputs the start and end data segments of the frame to be processed into low-pass filters respectively. The high-frequency components in the data segments are attenuated by the amplitude-frequency characteristics of the filters, resulting in smoother amplitude changes and more continuous phase transitions in the processed data segments. For example, this method can use an FIR (Finite Impulse Response) low-pass filter, performing convolution operations on the data segments using preset filter coefficients to achieve linear phase filtering and avoid introducing additional phase distortion during the filtering process. Alternatively, an IIR (Infinite Impulse Response) low-pass filter can be used to achieve high-frequency noise suppression with a simpler circuit or algorithm. Both filtering methods can achieve a smooth frequency band effect between the start and end data segments of the frame to be processed. The resulting smooth frequency band frame transition area can effectively eliminate the potential for signal abrupt changes when the frame to be processed connects with the preceding and following normal frames.

[0089] The frequency band smoothing frame transition area is covered to the missing frame position indicated by the whole frame missing flag to replace the data and obtain frame compensation data.

[0090] Specifically, based on the missing frame location information carried by the missing frame flag, the specific temporal interval corresponding to the missing frame in the dynamic buffer data is determined. This interval contains all the temporal positions of the sampling points that the missing frame should have, and is continuous with the temporal positions of the preceding and following normal frames. The previously obtained frequency band smoothing frame transition area is covered to the temporal interval of the missing frame according to the temporal correspondence: the frequency band smoothing data segment at the beginning of the frame to be processed covers the transition position between the beginning of the missing frame and the preceding normal frame, the frequency band smoothing data segment at the end of the frame to be processed covers the transition position between the end of the missing frame and the following normal frame, and the complete data segment in the middle of the frame to be processed that has not undergone filtering covers the middle temporal position of the missing frame. During the coverage process, the method replaces the original blank or abnormal data at the missing frame position with the coverage data through data replacement logic. The replacement process strictly follows the temporal alignment principle to ensure that the sampling point temporal of each coverage data is completely matched with the temporal of the corresponding position of the missing frame. Through the above steps, the frame compensation data obtained by this method not only retains the core features of the frame to be processed that are consistent with the original audio signal, but also eliminates the signal abrupt change at the connection point through the smooth connection area. It can completely fill the data gap caused by the missing whole frame and form a seamless transition with the normal frames before and after.

[0091] In one embodiment, a real-time audio data stream from a digital audio conferencing system is acquired, and the real-time audio data stream is input into a dynamic buffer using dynamic caching technology to obtain dynamic buffer data, including:

[0092] A circular buffer queue is used to store the real-time audio input stream using a sliding window, generating a continuously updated historical data set.

[0093] Specifically, this circular buffer queue employs a fixed-capacity storage structure (e.g., a hardware buffer built with dual-port RAM) with its write pointer moving cyclically with the continuous input of real-time audio data streams. For example, when 48kHz audio data is written at a rate of one sample point per clock cycle, a sliding window with a buffer coverage duration of 20ms dynamically generates a historical data set containing 960 consecutive sample points. In this design, read and write operations achieve zero-conflict parallel access through a dual-pointer mechanism with independent clock domains, ensuring the historical data set maintains temporal continuity during high-speed data stream input.

[0094] When new data is written, the oldest data that has exceeded the preset time limit is overwritten to generate a dynamic cache with a constant capacity.

[0095] Specifically, new sampling points cover the oldest data units in the buffer according to the circular address mapping rule, forming a first-in-first-out (FIFO) data update mechanism. For example, every time the write pointer increments by one address, the data at the original address is automatically erased and a new sampling point is written. At the same time, the pointer is reset to the beginning of the buffer through a modulo operation mechanism, generating a dynamic buffer with a constant capacity of 960 sampling points. This strategy differs from the static caching scheme, ensuring data real-time performance while avoiding linear growth of storage space.

[0096] Establish an index mapping table for the historical data set to access the Kth sampling point before the missing point and the most recent complete historical frame.

[0097] Specifically, to support efficient access to historical data, this technical solution establishes an index mapping table as the core addressing mechanism. For example, a bitmap index table can be constructed in the FPGA to establish a bidirectional mapping relationship between logical sampling point numbers (such as timing numbers 1 to 960) and physical memory addresses (such as RAM cells 0x0000 to 0x03C0). For example, when it is necessary to access the Kth sampling point before the missing point, the reference point logical number is calculated by a subtractor, and then the physical address translation and data reading are completed within one clock cycle. This hardware-level addressing scheme is faster than software traversal and provides low-latency reference data access capabilities for single-point compensation and frame copying.

[0098] In one embodiment, the compensation data is mapped to the original audio timing position in the real-time audio data stream for interpolation to obtain a continuous audio output stream, including:

[0099] Obtain the raw audio timing table of the real-time audio data stream. The raw audio timing table is used to record the timing position information of each sampling point in the real-time audio data stream. The timing position information includes the timestamp of the sampling point and its sequence number in the data stream.

[0100] Specifically, the original audio timing table of the real-time audio data stream is retrieved from the timing management module of the digital audio conferencing system. This timing table is established synchronously when the system generates the real-time audio data stream, and its data source is completely synchronized with the sampling process of the real-time audio data stream. Whenever the audio acquisition end completes the quantization encoding of a sampling point, the system automatically assigns a unique sequence number to that sampling point (to identify its order in the data stream) and records the timestamp when the sampling point was acquired (to identify its actual generation time). Subsequently, the sequence number and timestamp are associated and stored to form the original audio timing table. Furthermore, this method performs an integrity check on the retrieved original audio timing table to ensure that the timing position information of each sampling point in the table is complete and without errors, avoiding subsequent mapping deviations due to abnormal timing information.

[0101] Based on the missing type flag, the target temporal position corresponding to the missing data in the real-time audio data stream is determined. If the missing type flag is a single-point missing flag, the target temporal position is the temporal position of a single sampling point; if the missing type flag is a whole-frame missing flag, the target temporal position is the continuous temporal position interval corresponding to a frame of data.

[0102] Specifically, when determining the target temporal position based on the missing type flag, this technical solution employs a differentiated positioning strategy. If the missing type flag is a single-point missing flag, a subtractor is used to calculate the sequence number of the previous valid sampling point plus 1, and the temporal position corresponding to this value is used as the single compensation point insertion bit. If the missing type flag is a whole-frame missing flag, a continuous sequence number interval is calculated based on the frame length to generate a temporal position interval corresponding to the timestamp range. Furthermore, a hardware comparator array can be used to verify the continuity between the target position and adjacent temporal sequences in parallel, ensuring that the insertion operation does not disrupt the phase consistency of the original data stream.

[0103] The compensation data is inserted into the real-time audio data stream according to the target time sequence position, and the inserted data stream is interpolated to obtain a preliminary fused audio stream.

[0104] Specifically, this method first retrieves the corresponding compensation data based on the missing type flag. If it is a single-point missing data, the single-point compensation value is retrieved; if it is a whole-frame missing data, the frame compensation data is retrieved. Then, according to the target timing position determined by the original audio timing table, the compensation data is inserted one by one into the missing position of the real-time audio data stream: for single-point missing data, the single-point compensation value is directly inserted into the empty space corresponding to the target timing position; for whole-frame missing data, each sampling point of the frame compensation data is inserted into the continuous interval of the target timing position according to the sequence number. After insertion, the method detects the amplitude and phase connection status between the compensation data and the effective sampling points before and after the missing position. It finds that due to the difference in the generation timing between the compensation data and the original data, there may be slight amplitude jumps or phase breaks at the connection point. To address this issue, the method employs an interpolation algorithm adapted to the continuity of the audio signal: for single-point compensation, a linear interpolation algorithm is used to smoothly transition the compensation data with the preceding and following data based on the amplitude change rate of the effective sampling points before and after the missing point; for whole-frame compensation, a nonlinear interpolation algorithm (such as an interpolation algorithm based on the derivative of the audio signal) is used to gradually transition the start and end of the frame compensation data based on the amplitude and phase change trends of the frames before and after the missing frame, ensuring the overall continuity of the inserted data stream.

[0105] The initial fused audio stream is transmitted to a preset low-pass filter. The low-pass filter filters out abnormal high-frequency signals generated by data interpolation in the initial fused audio stream to obtain a continuous audio output stream. The abnormal high-frequency signals are signals that exceed the normal audio frequency range when the compensation data is connected to the original valid data.

[0106] For example, a preset low-pass filter in the digital audio conferencing system is retrieved. This filter is designed based on the normal audio frequency range of the conferencing scenario. Its passband frequency covers normal human voices and common effective audio frequencies in the conferencing environment, while its stopband frequency precisely targets abnormal high-frequency signals that may be generated during interpolation (these signals are usually outside the normal hearing range and have no effective audio information, only causing noise). Furthermore, an adaptive cutoff frequency control mechanism can be adopted to automatically configure the filter parameters according to the signal fundamental frequency: when the fundamental frequency of the voice is below 300Hz (e.g., male voice), a 3kHz cutoff frequency is set; when the fundamental frequency is above 500Hz (e.g., female voice), it is dynamically boosted to 5kHz to preserve overtone details.

[0107] In summary, the data loss compensation method for this digital audio conferencing system effectively addresses the issues of single-point data loss caused by clock source frequency differences and whole-frame data loss caused by transmission anomalies, thus improving sound abnormalities such as crackling and stuttering. Through a tiered loss detection mechanism, it accurately identifies single-point and whole-frame loss types and employs corresponding single-point compensation algorithms based on pre-stored sine wave signals and frame compensation algorithms based on the most recent complete historical frames and low-pass filtering. During the compensation process, it relies on a high-speed processing clock (with a processing rate far exceeding the audio sampling rate) and dynamic caching technology. This avoids the shortcomings of traditional compensation methods, such as filters failing to eliminate abnormal frequencies, fixed compensation differing from the current timbre, and significant delays caused by stored signals. It maximizes the restoration of the original characteristics of the audio signal, maintaining its continuity and consistent sound quality without adding extra delay, ensuring real-time audio transmission, and ultimately improving the audio output quality and user listening experience of the digital audio conferencing system in both wired and wireless scenarios.

[0108] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0109] Based on the same inventive concept, this application also provides a data loss compensation system 10 for a digital audio conferencing system to implement the data loss compensation method for the digital audio conferencing system described above. The solution provided by this system is similar to the implementation described in the above method. Therefore, the specific limitations of one or more embodiments of the data loss compensation system 10 for a digital audio conferencing system provided below can be found in the limitations of the data loss compensation method for a digital audio conferencing system described above, and will not be repeated here.

[0110] In one exemplary embodiment, such as Figure 2 As shown, a data loss compensation system 10 for a digital audio conferencing system is provided, comprising:

[0111] The audio data input module 11 is used to acquire the real-time audio data stream of the digital audio conferencing system and input the real-time audio data stream into the dynamic buffer area through dynamic caching technology to obtain dynamic buffer area data.

[0112] The data missing detection module 12 is used to perform missing type detection on the sampling point time interval of the dynamic buffer data and obtain missing type flags, wherein the missing type flags include single-point missing flags and / or whole-frame missing flags;

[0113] The data classification and compensation module 13 is used to determine the compensation strategy based on the missing type flag, and to compensate the digital audio signal based on the compensation strategy to obtain compensation data, which includes single-point compensation value and / or frame compensation data.

[0114] The data output module 14 is used to map the compensation data to the original audio timing position in the real-time audio data stream for interpolation to obtain a continuous audio output stream.

[0115] In one embodiment, the data missing detection module 12 includes:

[0116] The clock interval calculation unit is used to calculate the actual receiving interval based on the receiving timestamps of adjacent sampling points in the dynamic buffer data. The actual receiving interval is the arrival time difference between the current sampling point and the previous sampling point.

[0117] The single-point missing determination unit is used to determine a single-point missing status when the actual reception interval exceeds the preset standard sampling period, and obtain a single-point missing flag.

[0118] The frame missing conversion unit is used to count and accumulate the continuously obtained single-point missing flags through a counter. When the count value reaches the preset frame length, the frame missing conversion is performed to obtain the whole frame missing flag and the counter is reset.

[0119] In one embodiment, the data classification compensation module 13 includes:

[0120] The reference parameter extraction unit is used to extract the Nth valid sampling point before the missing point and its amplitude and phase characteristics from the dynamic buffer data based on the single-point missing flag, so as to obtain the waveform reconstruction reference parameters.

[0121] The phase synchronization processing unit is used to call the pre-stored sine wave signal template, perform phase synchronization processing based on the waveform reconstruction reference parameters, and obtain the phase-aligned sine wave template.

[0122] The single-point compensation value generation unit is used to adjust the amplitude of the phase-aligned sine wave template according to the amplitude change trend of the waveform reconstruction reference parameters to obtain the single-point compensation value.

[0123] In one embodiment, the phase synchronization processing unit includes:

[0124] The phase reference calculation subunit is used to calculate the instantaneous phase angle of the Nth valid sampling point before the missing point to obtain the current phase reference;

[0125] The synchronous sine wave generation subunit is used to perform offset compensation processing on the initial phase of the pre-stored sine wave signal template to generate a synchronous sine wave that is aligned with the current phase reference.

[0126] The phase adjustment amount in the offset compensation process is equal to that of the current phase reference.

[0127] In one embodiment, the data classification compensation module 13 further includes:

[0128] The frame data acquisition unit is used to copy the most recent complete historical frame from the dynamic buffer area based on the whole frame missing flag to obtain the frame data to be processed;

[0129] The smooth transition region generation unit is used to perform low-pass filtering on the start and end data segments of the frame data to be processed to obtain the frequency band smooth frame transition region.

[0130] The frame compensation data generation unit is used to cover the missing frame position indicated by the full frame missing flag with the frequency band smoothing frame transition area to replace the data and obtain frame compensation data.

[0131] In one embodiment, the audio data input module 11 includes:

[0132] The historical data set generation unit is used to perform sliding window storage processing on the real-time audio input stream through a circular buffer queue to generate a continuously updated historical data set.

[0133] The dynamic cache data generation unit is used to overwrite the oldest data that exceeds a preset time when new data is written, and generate dynamic cache data with a constant capacity.

[0134] The index mapping table creation unit is used to create an index mapping table for the historical data set, which is used to access the Kth sampling point before the missing point and the most recent complete historical frame.

[0135] In one embodiment, the data output module 14 includes:

[0136] The timing table acquisition unit is used to acquire the original audio timing table of the real-time audio data stream. The original audio timing table is used to record the timing position information of each sampling point in the real-time audio data stream. The timing position information includes the timestamp of the sampling point and its sequence number in the data stream.

[0137] The target temporal position determination unit is used to determine the target temporal position corresponding to the missing data in the real-time audio data stream based on the missing type flag. If the missing type flag is a single-point missing flag, the target temporal position is the temporal position of a single sampling point; if the missing type flag is an entire frame missing flag, the target temporal position is the continuous temporal position interval corresponding to a frame of data.

[0138] The preliminary fused audio stream generation unit is used to insert the compensation data into the real-time audio data stream according to the target time sequence position, and to perform interpolation on the inserted data stream to obtain the preliminary fused audio stream.

[0139] The continuous audio output stream generation unit is used to transmit the preliminary fused audio stream to a preset low-pass filter. The low-pass filter filters out abnormal high-frequency signals generated by data interpolation in the preliminary fused audio stream to obtain a continuous audio output stream. The abnormal high-frequency signals are signals that exceed the normal audio frequency range generated when the compensation data is connected to the original valid data.

[0140] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of a data loss compensation method for a digital audio conferencing system as described above.

[0141] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0142] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0143] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for compensating for data loss in a digital audio conferencing system, characterized in that, The method includes: The system acquires real-time audio data streams from a digital audio conferencing system and inputs these real-time audio data streams into a dynamic buffer using dynamic caching technology to obtain dynamic buffer data. Missing type detection is performed on the sampling point time interval of the dynamic buffer data to obtain missing type flags, wherein the missing type flags include single-point missing flags and / or whole-frame missing flags; A compensation strategy is determined based on the missing type flag, and the digital audio signal is compensated based on the compensation strategy to obtain compensation data, which includes single-point compensation values ​​and / or frame compensation data. The compensation data is mapped to the original audio timing position in the real-time audio data stream and interpolated to obtain a continuous audio output stream.

2. The method according to claim 1, characterized in that, The step of performing missing type detection on the sampling point time interval of the dynamic cache data to obtain a missing type flag includes: The actual receiving interval is calculated by clocking adjacent sampling points in the dynamic buffer data, where the actual receiving interval is the arrival time difference between the current sampling point and the previous sampling point. When the actual receiving interval exceeds the preset standard sampling period, a single point of absence determination is performed to obtain the single point of absence flag. The single-point missing flags obtained continuously are counted and accumulated by a counter. When the count value reaches the preset frame length, the frame missing flag is obtained by frame missing conversion and the counter is reset.

3. The method according to claim 1, characterized in that, The step of determining a compensation strategy based on the missing type flag and compensating the digital audio signal based on the compensation strategy to obtain compensation data includes: Based on the single-point missing flag, the Nth valid sampling point before the missing point and its amplitude and phase characteristics are extracted from the dynamic buffer data to obtain the waveform reconstruction reference parameters. A pre-stored sine wave signal template is invoked, and phase synchronization processing is performed based on the waveform reconstruction reference parameters to obtain a phase-aligned sine wave template. The amplitude of the phase-aligned sine wave template is adjusted according to the amplitude change trend of the waveform reconstruction reference parameters to obtain the single-point compensation value.

4. The method according to claim 3, characterized in that, The step of calling a pre-stored sine wave signal template and performing phase synchronization processing based on the waveform reconstruction reference parameters to obtain a phase-aligned sine wave template includes: The instantaneous phase angle is calculated for the Nth valid sampling point before the missing point to obtain the current phase reference; The initial phase of the pre-stored sine wave signal template is offset and compensated to generate a synchronous sine wave that is aligned with the current phase reference. The phase adjustment amount of the offset compensation process is equal to that of the current phase reference.

5. The method according to claim 1, characterized in that, The step of determining a compensation strategy based on the missing type flag and compensating the digital audio signal based on the compensation strategy to obtain compensation data further includes: Based on the missing frame flag, the most recent complete historical frame is copied from the dynamic buffer data to obtain the frame data to be processed; Low-pass filtering is performed on the start and end data segments of the frame data to be processed to obtain a smooth frame transition region. The frequency band smoothing frame transition area is covered to the missing frame position indicated by the whole frame missing flag to replace the data, thus obtaining the frame compensation data.

6. The method according to claim 1, characterized in that, The process of acquiring real-time audio data streams from a digital audio conferencing system involves inputting the real-time audio data streams into a dynamic buffer using dynamic caching technology to obtain dynamic buffer data, including: The real-time audio input stream is stored in a sliding window using a circular buffer queue to generate a continuously updated historical data set. When new data is written, the oldest data that exceeds the preset time is overwritten to generate the dynamic cache data with a constant capacity; An index mapping table is established for the historical data set to access the Kth sampling point before the missing point and the most recent complete historical frame.

7. The method according to claim 1, characterized in that, The step of mapping the compensated data to the original audio timing position in the real-time audio data stream for interpolation to obtain a continuous audio output stream includes: Obtain the original audio timing table of the real-time audio data stream. The original audio timing table is used to record the timing position information of each sampling point in the real-time audio data stream. The timing position information includes the timestamp of the sampling point and its sequence number in the data stream. Based on the missing type flag, the target temporal position corresponding to the missing data in the real-time audio data stream is determined. If the missing type flag is a single-point missing flag, the target temporal position is the temporal position of a single sampling point; if the missing type flag is an entire frame missing flag, the target temporal position is a continuous temporal position interval corresponding to a frame of data. The compensation data is inserted into the real-time audio data stream according to the target time sequence position, and the inserted data stream is interpolated to obtain a preliminary fused audio stream; The initial fused audio stream is transmitted to a preset low-pass filter. The low-pass filter filters out abnormal high-frequency signals generated by data interpolation in the initial fused audio stream to obtain the continuous audio output stream. The abnormal high-frequency signals are signals that exceed the normal audio frequency range generated when the compensation data is connected to the original valid data.

8. A data loss compensation system for a digital audio conferencing system, characterized in that, The system includes: The audio data input module is used to acquire the real-time audio data stream of the digital audio conferencing system and input the real-time audio data stream into the dynamic buffer area through dynamic caching technology to obtain dynamic buffer area data. The data missing detection module is used to perform missing type detection on the sampling point time interval of the dynamic buffer data and obtain a missing type flag, wherein the missing type flag includes a single point missing flag and / or an entire frame missing flag; The data classification and compensation module is used to determine a compensation strategy based on the missing type flag, and to compensate the digital audio signal based on the compensation strategy to obtain compensation data, which includes single-point compensation values ​​and / or frame compensation data. The data output module is used to map the compensation data to the original audio timing position in the real-time audio data stream for interpolation to obtain a continuous audio output stream.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images