A lossless coding and decoding method, device and medium for distributed audio and video processing

Abstract

The application provides a lossless coding and decoding method, device and medium for distributed audio and video processing, comprising: generating a preliminary node allocation scheme according to a frequency band grouping boundary, calculating the frequency band processing load of each node by using a load balancing algorithm, re-distributing the frequency band carried by the node whose load exceeds a preset load threshold to a low-load node to obtain a node allocation scheme; performing frequency band splitting on multi-channel audio according to the node allocation scheme, driving each node to independently encode the allocated frequency band to obtain a node frequency band signal; performing cross-node phase synchronization detection on each node frequency band signal, extracting a node pair with different phases and the corresponding frequency band, re-distributing the frequency band to a phase-coherent node to obtain an optimized node allocation scheme; performing distributed coding on each node frequency band signal by using the optimized node allocation scheme and recombining, extracting the spatial sound field positioning of the recombined signal, comparing with the original signal positioning, and identifying the final lossless audio signal meeting the preset positioning range.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a lossless encoding and decoding method, device and medium for distributed audio and video processing. Background Technology

[0002] In the field of audio and video processing, research on distributed systems is of great significance for improving the encoding and restoration quality of multi-channel audio. This field is directly related to the realism of the user's immersive auditory experience, especially in scenarios such as virtual reality and cinema sound effects. Ensuring the spatial sense and positioning accuracy of sound has become a core goal of technological development. With the widespread application of distributed technologies, the processing efficiency and quality of audio signals have been significantly improved. However, how to maintain the original characteristics of sound in complex environments remains a challenge that urgently needs to be addressed.

[0003] Currently, many solutions tend to split multi-channel audio signals into different frequency ranges and distribute them across multiple processing nodes to maximize resource utilization. However, this approach often overlooks the inherent connections between different frequency ranges of the sound signal. Especially in a distributed environment, when each node processes the signal independently, the lack of overall coordination can lead to compromised signal characteristics. This deficiency is not simply a technical implementation problem, but stems from insufficient attention to the overall characteristics of the sound signal. Particularly in cross-node processing, certain key attributes of the signal are easily fragmented, and the phase relationship between different frequency components of the sound signal becomes a crucial factor affecting spatial perception. Phase relationship refers to the temporal synchronicity of the sound signal. If the phases of the low-frequency and high-frequency components are inconsistent, the spatial positioning of the sound will be inaccurate.

[0004] Especially in distributed systems, when audio signals are split into multiple frequency bands and distributed to different nodes for processing, the independent encoding by each node may lead to phase shifts. These shifts accumulate during signal reconstruction, ultimately disrupting the spatial localization of the sound. For example, in a multi-channel surround sound scenario, suppose a sound should come from the listener's left, but due to phase shifts, the low and high frequencies cannot be synchronized after reconstruction. The listener may misinterpret the sound as coming from the front or the right, and this misalignment directly affects the immersive experience.

[0005] Therefore, in distributed audio processing, how to balance the efficiency of signal splitting and node allocation while avoiding spatial positioning deviations caused by phase relationship damage has become a key issue in improving the user's auditory experience. Summary of the Invention

[0006] This invention provides a lossless encoding and decoding method for distributed audio and video processing, comprising:

[0007] Obtain the spatial coordinates of the physical loudspeakers in each channel, construct a channel topology diagram, extract the coherence level between channels based on the channel topology diagram, mark the frequency bands with coherence levels below a preset threshold as phase-sensitive frequency bands, and generate a combination of phase-sensitive frequency bands.

[0008] Based on the combination of phase-sensitive frequency bands, the phase-locking state of the low-frequency and high-frequency components of the same channel is analyzed, and a preliminary node allocation scheme is generated based on the frequency band grouping boundary.

[0009] Based on the node allocation scheme, the multi-channel audio is split into frequency bands, and cross-node phase synchronization detection is performed on the frequency band signals of each node;

[0010] An optimized node allocation scheme is used to perform distributed coding and reassembly on the frequency band signals of each node.

[0011] Furthermore, the step of extracting the inter-channel coherence levels, marking frequency bands with coherence levels below a preset threshold as phase-sensitive frequency bands, and generating a combination of phase-sensitive frequency bands specifically includes:

[0012] Traverse each pair of adjacent channel nodes in the channel topology diagram, extract the audio signals of adjacent channels within the same time window, perform a short-time Fourier transform on the audio signals to obtain the amplitude and phase of each frequency component, multiply the frequency components of two adjacent channels point by point for each frequency band, sum and take the average value to obtain the coherence value, if the coherence value is lower than the preset coherence threshold, then mark the frequency band as a phase-sensitive frequency band;

[0013] The number of channel pairs marked as phase-sensitive frequency bands in the same frequency band in the channel topology diagram is counted, and the frequency bands whose number of channel pairs exceeds a preset threshold are classified into phase-sensitive frequency band combinations. The phase-sensitive frequency band combination includes the frequency range of each sensitive frequency band and its corresponding channel pair association information.

[0014] Furthermore, the step of analyzing the phase-locking state of the low-frequency and high-frequency components of the same channel based on the combination of phase-sensitive frequency bands specifically includes:

[0015] The audio signal within the same channel is extracted based on the phase-sensitive frequency band combination. The audio signal is then subjected to a short-time Fourier transform to obtain a time-frequency representation. The signal is divided into low-frequency components and high-frequency components according to their frequency. The instantaneous phase values ​​of the low-frequency components and high-frequency components at the same time point are extracted to obtain a low-frequency phase sequence and a high-frequency phase sequence.

[0016] The phase difference between the low-frequency phase sequence and the high-frequency phase sequence at the same time point is calculated. The change amplitude of the phase difference over a continuous time period is statistically analyzed. If the change amplitude is lower than a preset stability threshold, the low-frequency component and the high-frequency component in the corresponding time period are determined to be in a phase-locked state. If the change amplitude exceeds the preset stability threshold, the time period is determined to be in a phase-unlocked state. The frequency position corresponding to the phase-unlocked state is recorded as the unlocked position.

[0017] Furthermore, the method also includes:

[0018] Based on the distribution of the unlocked positions on the frequency axis, the unlocked positions are used as candidate points for the frequency band group boundary. For each sensitive frequency band combination, the number of unlocked positions of its low-frequency components and the number of unlocked positions of its high-frequency components are counted. The unlocked ratio is obtained by dividing the number of unlocked positions of the low-frequency components by the number of unlocked positions of the high-frequency components.

[0019] If the unlock ratio exceeds the preset allowable threshold, the sensitive frequency band combination is merged with the adjacent sensitive frequency band combination on the frequency axis. The unlock position is re-detected and the boundary is delineated within the merged frequency band range to obtain the frequency band group boundary.

[0020] Furthermore, the step of generating a preliminary node allocation scheme based on frequency band grouping boundaries specifically includes:

[0021] Based on the frequency band grouping boundary, obtain the frequency range and bandwidth information of each frequency band, extract the list of available nodes from the distributed processing cluster, establish a frequency band node mapping table according to the correspondence between the number of frequency bands and the number of nodes, the frequency band node mapping table records the target node identifier and the bandwidth value of each frequency band, and sequentially allocate each frequency band within the frequency band grouping boundary to each node in the node list to obtain a preliminary node allocation scheme;

[0022] According to the preliminary node allocation scheme, each node is traversed, and the bandwidth value of the frequency band carried by each node is accumulated as the processing load of that node. The processing load is compared with a preset high load threshold. If the processing load exceeds the preset high load threshold, the node is marked as a high load node. If the processing load is lower than the preset low load threshold, the node is marked as a low load node.

[0023] For the frequency bands carried by the high-load nodes, the frequency bands to be migrated are extracted one by one in descending order of the bandwidth values ​​recorded in the frequency band node mapping table. The node with the smallest current processing load is selected from the low-load nodes as the receiving node. The frequency bands to be migrated are migrated from the high-load nodes to the receiving nodes, and the node identifiers in the frequency band node mapping table are updated to obtain the node allocation scheme.

[0024] Furthermore, the frequency band splitting of multi-channel audio based on the node allocation scheme specifically includes:

[0025] According to the node allocation scheme, obtain the frequency band list carried by each node, perform bandpass filtering on each channel signal of the multi-channel audio according to the frequency range recorded in the frequency band list, extract the sub-band signal of the corresponding frequency band from each channel signal, and distribute the sub-band signal to the corresponding node according to the node identifier to obtain the sub-band signal to be encoded for each node.

[0026] For the sub-band signal to be encoded at each node, each node is driven to perform independent encoding. The independent encoding performs quantization and entropy encoding on the sub-band signal. After each node completes independent encoding, it outputs the corresponding encoded bit stream as the frequency band signal of each node, thus obtaining the frequency band signal of each node.

[0027] Furthermore, the cross-node phase synchronization detection of signals in each node frequency band specifically includes:

[0028] The phase information of the frequency band signal carried by each node is obtained. For every two adjacent nodes, a node pair is constructed. For each frequency band, the phase sequence of the two nodes in the node pair in that frequency band is extracted. The phase difference is calculated point by point in time. The standard deviation of the phase difference is calculated. If the standard deviation of the phase difference is lower than the preset synchronization threshold, the node pair is phase synchronized. Otherwise, it is marked as a phase-asynchronous node pair. The frequency band identifier shared by the two nodes in the phase-asynchronous node pair is extracted. The frequency band identifier is associated with the corresponding node pair information and recorded to obtain a list of asynchronous frequency bands.

[0029] According to the list of asynchronous frequency bands, each pair of asynchronous nodes is traversed. For each asynchronous frequency band, the phase sequence of other nodes in the current node allocation scheme is retrieved from the list. The standard deviation between the phase sequence of the relevant node in the frequency band and the original phase sequence of the frequency band is calculated. If the standard deviation is lower than the preset synchronization threshold, it is marked as a phase coherent node.

[0030] The asynchronous frequency bands are migrated from the original bearer nodes to the phase coherent nodes, and the frequency band and node mapping relationship in the node allocation scheme is updated to obtain an optimized node allocation scheme.

[0031] Furthermore, the step of performing distributed coding and reassembly on the frequency band signals of each node using an optimized node allocation scheme specifically includes:

[0032] For the reconstructed audio signal, the amplitude ratio between each channel is extracted as the sound level difference, and the arrival time difference between each channel is extracted as the time difference. Based on the sound level difference and the time difference, the azimuth and elevation angles of the sound source in three-dimensional space are calculated. The azimuth angle is calculated as the sound speed multiplied by the time difference divided by the arcsine of the head width. The elevation angle is calculated as the sound level difference divided by the arctangent of an empirical constant. The azimuth and elevation angles are combined to form the spatial sound field positioning of the reconstructed signal.

[0033] The spatial sound field positioning of the original multi-channel audio signal is extracted in the same way as the reconstructed signal and used as the original positioning reference. The spatial sound field positioning of the reconstructed signal is compared with the original positioning reference channel by channel, and the difference between the two in azimuth and elevation angles is calculated to obtain the positioning deviation.

[0034] Secondly, this invention discloses a lossless encoding and decoding device for distributed audio and video processing, the device comprising:

[0035] The memory, the processor, and the computer program stored on the memory and executable on the processor, the computer program being configured to implement the lossless encoding and decoding steps of a distributed audio and video processing as described above.

[0036] Thirdly, the present invention discloses a medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the aforementioned lossless encoding and decoding method for distributed audio and video processing.

[0037] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0038] This invention discloses a lossless encoding and decoding method for distributed audio and video processing, proposing a complete solution for complex business scenarios involving frequency band allocation, phase synchronization, and load balancing in multi-channel audio processing. The core challenge lies in ensuring the synchronization of phase-sensitive frequency bands and the accuracy of spatial sound field localization while simultaneously avoiding uneven node load, all while achieving frequency band splitting and encoding among distributed nodes. This invention constructs a channel topology diagram, extracts coherence levels, marks phase-sensitive frequency bands, and dynamically adjusts the node allocation scheme by combining frequency band grouping boundary optimization and load balancing algorithms to ensure balanced processing load across nodes. Furthermore, through cross-node phase synchronization detection and distributed encoding reassembly, this invention effectively solves the phase lock-out problem, ultimately outputting a lossless audio signal with a high degree of consistency with the original signal localization. The overall technical effect is a significant improvement in the encoding efficiency and sound field restoration accuracy of multi-channel audio processing, providing innovative assurance for distributed audio and video processing. Attached Figure Description

[0039] Figure 1 This is a flowchart of a lossless encoding and decoding method for distributed audio and video processing according to the present invention.

[0040] Figure 2 This is a schematic diagram of a lossless encoding and decoding method for distributed audio and video processing according to the present invention.

[0041] Figure 3 This is another schematic diagram of a lossless encoding and decoding method for distributed audio and video processing according to the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0043] like Figures 1-3 This embodiment of a lossless encoding / decoding method, device, and medium for distributed audio and video processing may specifically include:

[0044] Step S101: Obtain the spatial coordinates of the physical loudspeakers in each channel, construct a channel topology diagram, extract the coherence level between channels based on the channel topology diagram, mark the frequency bands with coherence levels below a preset threshold as phase-sensitive frequency bands, and generate a combination of phase-sensitive frequency bands.

[0045] The system acquires the position information of each physical speaker in three-dimensional space for multi-channel audio, extracts the horizontal azimuth and vertical elevation angles of each speaker, calculates the angular spacing between adjacent channels based on these angles, and uses these angular spacings as edge weights in the channel topology graph. An adjacency matrix representation is used to construct the channel topology graph, where each node corresponds to a physical speaker, and the connections between nodes reflect the spatial adjacency of the channels. Each pair of adjacent channel nodes in the channel topology graph is traversed, and the audio signals of adjacent channels within the same time window are extracted. A short-time Fourier transform is performed on the audio signals to obtain the amplitude and phase of each frequency component. For each frequency band, the frequency components of two adjacent channels are multiplied point-by-point, summed, and averaged to obtain the coherence value. If the coherence value is lower than a preset coherence threshold, the frequency band is marked as a phase-sensitive band. The number of channel pairs marked as phase-sensitive frequency bands in the same frequency band in the channel topology diagram is counted. Frequency bands with more than a preset threshold number of channel pairs are classified into phase-sensitive frequency band combinations. The phase-sensitive frequency band combination includes the frequency range of each sensitive frequency band and its corresponding channel pair association information, thus obtaining the phase-sensitive frequency band combination.

[0046] For example, taking a seven-channel surround sound configuration as an example, each speaker is distributed in different positions around the listener. The spatial coordinates of the physical speakers are the basic data source for constructing the channel topology diagram, including the front left channel, front right channel, center channel, surround left channel, surround right channel, rear left channel, and rear right channel. The position of each speaker in three-dimensional space is represented by the horizontal azimuth angle and the vertical elevation angle. The horizontal azimuth angle is based on the zero-degree reference directly in front of the listener, with the clockwise direction being the positive angle increment. The vertical elevation angle is based on the horizontal plane as the zero-degree reference, with the upward direction being the positive angle increment.

[0047] Specifically, the channel topology diagram is represented in the form of an adjacency matrix, where the rows and columns correspond to each channel node, and the values ​​of the matrix elements are the angular distances between adjacent channels.

[0048] In one embodiment, the horizontal azimuth angle of the front left channel is -30 degrees, and the horizontal azimuth angle of the front right channel is +30 degrees, with an angular interval of 60 degrees between them. This value serves as the edge weight for the corresponding position in the adjacency matrix. If two channels are not spatially adjacent, the corresponding element in the adjacency matrix is ​​set to infinity or a specific identifier value, indicating that there is no direct topological connection between them. The channel topology diagram constructed in this way can fully represent the spatial adjacency relationships between each channel, providing a structured traversal path for subsequent coherence calculations.

[0049] In one possible implementation, the coherence value is calculated based on the frequency domain analysis results of the short-time Fourier transform. The short-time Fourier transform divides the audio signal into several overlapping time windows. A Fourier transform is performed on the signal within each time window to obtain the amplitude and phase information of each frequency component within that time period. For each frequency band, let the complex frequency component sequences of two adjacent channels within the same time window be X(f,t) and Y(f,t), respectively. The two sequences are multiplied point-by-point by their conjugates, and the mean is calculated along the time window direction. The magnitude of the resulting mean is the coherence value for that frequency band, calculated using the following formula: ,in The coherence is the conjugate of Y(f,t), mean(·) represents the mean operation along the time window direction, and COH(f) takes values ​​in the range [0,1]. If the coherence value COH(f) is lower than the preset coherence threshold T, coh (The typical value is 0.7, which is determined based on the lower limit of human hearing's perception of phase consistency and can be calibrated according to actual sound field test data), then this frequency band is marked as a phase-sensitive frequency band.

[0050] It should be noted that the coherence value ranges between zero and one. The closer the value is to one, the higher the synchronization degree of the signals of two adjacent channels in that frequency band. The closer the value is to zero, the more random the phase relationship between the two signals is. When the coherence value of a certain frequency band is lower than the preset coherence threshold, it indicates that the frequency band is prone to phase shift due to independent encoding by nodes during distributed processing. Therefore, it is marked as a phase-sensitive frequency band.

[0051] Preferably, after calculating the coherence of all adjacent channel pairs, the number of times the same frequency band is marked as a phase-sensitive frequency band in different channel pairs is counted. If the proportion of the number of times a certain frequency band is marked as a phase-sensitive frequency band to the total number of channel pairs exceeds a preset proportion threshold T, the calculation is performed. pair (A typical value is 0.5, meaning that when more than half of the channel pairs exhibit phase-sensitive characteristics in this frequency band, it is included in the group; this can be adjusted according to the actual channel configuration.) Then, this frequency band is included in the phase-sensitive frequency band group. The phase-sensitive frequency band group records the start frequency, cutoff frequency, and associated channel pair information of each sensitive frequency band, providing a basis for subsequent frequency band grouping boundary delineation.

[0052] Step S102: Based on the phase-sensitive frequency band combination, analyze the phase-locking status of the low-frequency and high-frequency components of the same channel, determine the phase-locking position as the frequency band grouping boundary, and merge and redefine the boundary for sensitive frequency band combinations where the ratio of the number of low-frequency component lock-out positions to the number of high-frequency component lock-out positions exceeds a preset allowable threshold to obtain the frequency band grouping boundary.

[0053] Audio signals within the same channel are extracted based on phase-sensitive frequency band combinations. A short-time Fourier transform is performed on the audio signals to obtain a time-frequency representation, which is then divided into low-frequency and high-frequency components. The instantaneous phase values ​​at the same time point are extracted for each of the low-frequency and high-frequency components, resulting in low-frequency and high-frequency phase sequences. The phase difference between the low-frequency and high-frequency phase sequences at the same time point is calculated, and the amplitude of the phase difference over a continuous time period is statistically analyzed. If the amplitude is below a preset stability threshold, the low-frequency and high-frequency components are determined to be in a phase-locked state during that time period; if the amplitude exceeds the preset stability threshold, the time period is determined to be in a phase-unlocked state. The frequency position corresponding to the phase-unlocked state is recorded as the unlock position. Based on the distribution of the unlock positions on the frequency axis, these unlock positions are used as candidate points for frequency band grouping boundaries. For each sensitive frequency band combination, the number of unlock positions for the low-frequency and high-frequency components is statistically analyzed, and the unlock ratio is obtained by dividing the number of unlock positions for the low-frequency components by the number of unlock positions for the high-frequency components. If the unlock ratio exceeds the preset allowable threshold, the sensitive frequency band combination is merged with the adjacent sensitive frequency band combination on the frequency axis. The unlock position is re-detected and the boundary is delineated within the merged frequency band range to obtain the frequency band group boundary.

[0054] In distributed processing scenarios for multi-channel audio, the audio signal within the same channel contains a complete spectrum of components from low to high frequencies. The phase relationship between the low-frequency components and the high-frequency components directly affects the spatial positioning accuracy of the sound.

[0055] For example, in a surround sound system, low-frequency components typically carry the basic energy and directional information of the sound, while high-frequency components carry the detailed texture and presence information of the sound. The degree of synchronization between the two in time determines the accuracy of the listener's judgment of the sound source location.

[0056] Specifically, the division between low-frequency and high-frequency components is based on the spectral characteristics of the audio signal of the vocal tract. After extracting the audio data of the target vocal tract from the combination of phase-sensitive frequency bands, it is separated into two independent signal parts according to the frequency.

[0057] In one embodiment, the selection of the dividing frequency is related to the range of frequencies at which the human ear is sensitive to spatial localization. Signals below this dividing frequency are classified as low-frequency components, and signals above this dividing frequency are classified as high-frequency components. A typical dividing frequency is 800Hz, which corresponds to the upper limit of the effective frequency for the human ear to locate sound sources in the horizontal direction using phase difference (below 800Hz, the interaural time difference (ITD) is the main localization cue). It can be adjusted within the range of 500Hz to 1000Hz according to the actual sound field configuration.

[0058] In one possible implementation, the instantaneous phase is extracted using the Hilbert transform method. The Hilbert transform converts the real-valued audio signal into an analytic signal. The real part of the analytic signal is the original signal, and the imaginary part is the Hilbert transform result of the original signal. The instantaneous phase value can be obtained by calculating the argument of the analytic signal at each sampling point. After performing the Hilbert transform on the low-frequency components, the argument value is calculated point by point to form a low-frequency phase sequence; the same operation is performed on the high-frequency components to form a high-frequency phase sequence. The low-frequency phase sequence and the high-frequency phase sequence have the same time axis scale, allowing subsequent phase difference calculations to be performed sequentially at corresponding time points.

[0059] It should be noted that determining the phase-locked state is a crucial step in identifying frequency band group boundaries in this scheme. After obtaining the low-frequency and high-frequency phase sequences, the phase values ​​at the same time point are subtracted to obtain the phase difference sequence. If the low-frequency and high-frequency components maintain a stable phase relationship within a certain time period, the phase difference sequence within that time period exhibits stable characteristics, with its numerical fluctuations remaining within a small range; if the phase relationship between the two components drifts or jumps, the phase difference sequence exhibits drastic fluctuation characteristics. By setting a preset stability threshold as a determination threshold, when the change in phase difference within a continuous time period is lower than the determination threshold, it is determined that the low-frequency and high-frequency components are in a phase-locked state within that time period; when the change exceeds the determination threshold, it is determined that the time period is in a phase-unlocked state. In this embodiment, the preset stability threshold T... stable A typical value for this is π / 6 rad (approximately 30°). This value is based on the human ear's perception threshold for phase jumps. When the change in phase difference over a continuous time period is less than T... stable When the state is determined to be phase locked, it can be adjusted within the range of π / 12 to π / 4 according to the actual coding accuracy requirements.

[0060] Preferably, the statistical analysis of the phase difference variation amplitude is performed using a sliding time window. Within each sliding window, the range or standard deviation of the phase difference is calculated as a quantitative indicator of the variation amplitude. The length of the sliding window is set to a multiple of the audio signal sampling rate, typically the number of sampling points corresponding to a 20ms audio frame (e.g., 960 sampling points for a sampling rate of 48kHz). The movement step is set to 50% of the window length (i.e., 480 sampling points), and can be adjusted according to real-time processing latency requirements. The sliding window moves along the time axis, detecting the distribution of phase-locked and phase-unlocked states segment by segment. The time period corresponding to the phase-unlocked state is then mapped to a specific position on the frequency axis, forming a record of the unlocked position.

[0061] In one embodiment, the mapping of the unlocked position to the frequency axis is achieved based on the time-frequency correspondence of the short-time Fourier transform. When a certain time period is determined to be in a phase unlocked state, the frequency component information corresponding to that time period is extracted, and the center frequency of the frequency component is recorded as the unlocked position on the frequency axis. Multiple unlocked positions are discretely distributed on the frequency axis, and the frequency intervals between adjacent unlocked positions constitute independent frequency band grouping candidate regions.

[0062] Understandably, the lockout ratio is used to assess the difference in phase stability between low-frequency and high-frequency components within the same sensitive frequency band combination. The number of lockout locations for low-frequency components within the sensitive frequency band combination is counted, as are the number of lockout locations for high-frequency components within the same range. The former is divided by the latter to obtain the lockout ratio. When the lockout ratio exceeds a preset allowable threshold, it indicates that the phase stability of the low-frequency portion of the sensitive frequency band combination is significantly weaker than that of the high-frequency portion, making it more prone to phase shift accumulation during independent coding by distributed nodes. A typical value for the preset allowable threshold, Tratio, is 2.0. This means that when the number of lockout locations for low-frequency components exceeds twice the number of lockout locations for high-frequency components, the sensitive frequency band combination is considered to need to be merged. This value is based on the empirical rule that the phase stability of low-frequency components is generally weaker than that of high-frequency components and can be calibrated according to the actual frequency band distribution characteristics. Furthermore, for sensitive frequency band combinations with lockout ratios exceeding the allowable threshold, they are merged with adjacent sensitive frequency band combinations on the frequency axis. The merging operation combines two or more adjacent sensitive frequency bands into a wider frequency band range. Within this merged frequency band range, the loss-of-lock position detection is re-executed, and the frequency band grouping boundary is defined based on the newly detected loss-of-lock position, so that the final frequency band grouping boundary can take into account the phase characteristic differences between low-frequency and high-frequency components.

[0063] Step S103: Generate a preliminary node allocation scheme based on the frequency band grouping boundary, use a load balancing algorithm to calculate the frequency band processing load of each node, and reallocate the frequency bands carried by nodes whose load exceeds the preset load threshold to low-load nodes to obtain the node allocation scheme.

[0064] Based on the frequency band grouping boundaries, the frequency range and bandwidth information of each frequency band are obtained. An available node list is extracted from the distributed processing cluster. A frequency band node mapping table is established according to the correspondence between the number of frequency bands and the number of nodes. The frequency band node mapping table records the target node identifier and bandwidth value allocated to each frequency band. Each frequency band within the frequency band grouping boundary is sequentially allocated to the nodes in the node list to obtain a preliminary node allocation scheme. Each node is traversed according to the preliminary node allocation scheme, and the bandwidth value of the frequency band carried by each node is accumulated as the processing load of that node. The processing load is compared with a preset high-load threshold. If the processing load exceeds the preset high-load threshold, the node is marked as a high-load node; if the processing load is lower than the preset low-load threshold, the node is marked as a low-load node. For the frequency bands carried by the high-load nodes, the frequency bands to be migrated are extracted one by one in descending order of the bandwidth values ​​recorded in the frequency band node mapping table. The node with the smallest current processing load is selected from the low-load nodes as the receiving node. The frequency bands to be migrated are migrated from the high-load nodes to the receiving nodes, and the node identifiers in the frequency band node mapping table are updated to obtain the node allocation scheme.

[0065] In distributed audio processing scenarios, frequency band grouping boundaries define the frequency range of each independent frequency band, and each frequency band corresponds to a certain bandwidth value.

[0066] For example, in a multi-channel surround sound processing task, the frequency band grouping boundary divides the entire audio spectrum into several non-overlapping frequency band intervals, and the bandwidth value of each frequency band interval reflects the number of frequency components contained in that frequency band and the processing complexity.

[0067] Specifically, the frequency band node mapping table is a data structure that records the correspondence between frequency bands and processing nodes. Each record in the table contains the frequency band identifier, the start frequency, the cutoff frequency, the bandwidth value, and the identifier of the assigned target node.

[0068] In one embodiment, the distributed processing cluster includes multiple computing nodes, each with independent audio encoding processing capabilities. During the initial allocation, frequency bands are sequentially allocated to each node according to their arrangement, so that each node carries several consecutive or non-consecutive frequency bands.

[0069] It should be noted that the processing load is calculated by summing the bandwidth values ​​of the frequency bands carried by the node. The larger the bandwidth value of a frequency band, the more computational resources it consumes during encoding processing. Therefore, the processing load of a node can be quantified by summing the bandwidth values ​​of all frequency bands carried by the node.

[0070] In one possible implementation, all records belonging to the same node in the frequency band node mapping table are traversed, and the bandwidth values ​​in these records are summed one by one to obtain the processing load value of that node. For node i, let the set of frequency bands it carries be F. i The bandwidth value for each frequency band is BW. j (j∈F) i The node's load is handled according to the formula. Calculate, where BW j The unit is Hz, L i The unit is Hz, reflecting the total frequency bandwidth carried by the node.

[0071] Preferably, the high-load threshold and low-load threshold are set based on the ratio between the overall processing capacity of the distributed cluster and the total bandwidth of the frequency band. Let the total number of cluster nodes be N, the total audio bandwidth be BWtotal, and the average load be BW. avg =BW total / N, High load threshold T high =1.3×BW avg (i.e., exceeding 30% of the average load is considered a high load), low load threshold T low =0.7×BW avg(i.e., a load below 30% of the average load is considered low load). The proportional coefficients of 1.3 and 0.7 are based on the commonly used tolerance range of distributed load balancing and can be adjusted according to the actual performance differences of the cluster. When the processing load of a node exceeds the high load threshold, it indicates that the node is carrying too heavy a frequency band processing task, which is prone to processing delays during actual encoding. When the processing load of a node is below the low load threshold, it indicates that the node's computing resources are not being fully utilized and it has the ability to receive additional frequency bands.

[0072] In one embodiment, the frequency band migration follows a bandwidth-first principle, selecting frequency bands with larger bandwidth values ​​from those carried by high-load nodes as the migration targets. Prioritizing the migration of frequency bands with larger bandwidth values ​​enables rapid load balancing within fewer migration operations. The selection of receiving nodes follows a load-minimizing principle, choosing the node with the smallest current processing load value from low-load nodes as the receiver. After the migration is completed, the node identification information of the corresponding frequency band in the frequency band node mapping table is updated to form the final node allocation scheme.

[0073] Step S104: Perform frequency band splitting on the multi-channel audio according to the node allocation scheme, drive each node to independently encode the allocated frequency band, and obtain the frequency band signal of each node.

[0074] According to the node allocation scheme, the frequency band list carried by each node is obtained. For each channel signal of the multi-channel audio, bandpass filtering is performed according to the frequency range recorded in the frequency band list. The sub-band signal of the corresponding frequency band is extracted from each channel signal. The sub-band signal is distributed to the corresponding node according to the node identifier, resulting in the sub-band signal to be encoded for each node. For the sub-band signal to be encoded for each node, each node is driven to perform independent encoding. The independent encoding performs quantization and entropy encoding processing on the sub-band signal. After completing the independent encoding, each node outputs the corresponding encoded bitstream as the frequency band signal of each node, resulting in the frequency band signal of each node.

[0075] In distributed audio processing scenarios, the node allocation scheme records the frequency band information carried by each processing node, including the start and end frequencies of the frequency band.

[0076] For example, the bandpass filter performs filtering operations on the channel signals according to the frequency range recorded in the frequency band list, retaining only the signal components that fall within the specified frequency range and filtering out the frequency components outside the range, thereby extracting the sub-band signal corresponding to the frequency band.

[0077] Specifically, subband signals are distributed according to node identifiers, and subband signals belonging to the same node are packaged and sent to that node's processing queue.

[0078] In one embodiment, each node in the distributed processing cluster initiates an independent encoding process after receiving the subband signal, and each node processes the frequency band data it carries without interfering with each other.

[0079] It should be noted that independent coding comprises two stages: quantization and entropy coding. Quantization maps the continuous amplitude values ​​x of the subband signal to discrete quantization levels; the quantization formula is x... q =round(x / Δ)×Δ, where Δ is the quantization step size. A max B is the maximum absolute value of the subband signal amplitude, B is the number of quantization bits (typically 16 bits, which can be adjusted from 8 to 24 bits according to the target audio quality and bit rate requirements), and round(·) represents the rounding operation; entropy coding encodes the quantized discrete data xq using a lossless compression method (such as Huffman coding or arithmetic coding), and each node outputs the encoded bit stream after completing the encoding.

[0080] Step S105: Perform cross-node phase synchronization detection on the frequency band signals of each node, extract the node pairs with phase asynchrony and their corresponding frequency bands, and redistribute the frequency bands to phase coherent nodes to obtain an optimized node allocation scheme.

[0081] The process involves acquiring the phase information of the frequency band signals carried by each node, constructing a node pair for every two adjacent nodes, extracting the phase sequences φ1(t) and φ2(t) of the two nodes in the node pair within that frequency band, calculating the phase difference Δφ(t) = φ1(t) - φ2(t) at each time point, and calculating the standard deviation std(Δφ) of Δφ(t). If std(Δφ) < ​​5°, the node pair is considered phase-synchronized; otherwise, it is marked as a phase-asynchronous node pair. The frequency band identifier shared by the two nodes in the phase-asynchronous node pair is extracted, and the frequency band identifier is associated with the corresponding node pair information to obtain a list of phase-asynchronous frequency bands. Based on the list of phase-asynchronous frequency bands, each phase-asynchronous node pair is traversed. For each phase-asynchronous frequency band, the phase sequences of other nodes in that frequency band are retrieved from the current node allocation scheme. The std(Δφ) between the phase sequence of the candidate node in that frequency band and the original phase sequence of that frequency band is calculated. If std(Δφ) < ​​5°, it is marked as a phase-coherent node. The asynchronous frequency bands are migrated from the original bearer nodes to the phase coherent nodes, and the frequency band and node mapping relationship in the node allocation scheme is updated to obtain an optimized node allocation scheme.

[0082] In distributed audio processing scenarios, there is a phase relationship between the frequency band signals generated by each node after independent encoding. Cross-node phase synchronization detection is a key step in evaluating the consistency of the encoding results of each node.

[0083] For example, the construction of node pairs is based on the frequency range carried by the nodes on the frequency axis. When the frequency bands carried by two nodes are adjacent or overlap on the frequency axis, these two nodes are paired for phase comparison. If the frequency bands do not overlap but the interval is less than the threshold d (d=0.5MHz), then the edge frequency sequences of the two frequency bands are input, and a virtual overlapping sequence of length 10 is generated using a linear interpolation algorithm as the same frequency band range, and the phase comparison sequence is output. Specifically, a virtual overlapping sequence is generated between the edge frequencies of the two frequency bands using a linear interpolation algorithm, and the sequence length N is... interp According to formula N interp =round(d / f res ) calculate, where d is the frequency spacing between the two frequency bands (Hz), f res The frequency resolution (Hz / point, determined by the short-time Fourier transform window length) is typically around 10 sampling points (with d=0.5MHz, f res (For example, 50kHz / point), the resolution can be adjusted according to the actual frequency resolution.

[0084] Specifically, phase sequence extraction is performed on the signals of the two nodes in the node pair within the same frequency band. The coded bitstream of a specified frequency band is extracted from the frequency band signal of each node. This bitstream is decoded to reconstruct the sub-band signal. Then, the instantaneous phase values ​​of the sub-band signal at each sampling point are extracted using Hilbert transform, forming the phase sequence of that node in that frequency band. After the two nodes in the node pair have extracted their phase sequences, a time-point phase comparison can be performed.

[0085] It should be noted that the calculation of node phase synchronization is based on the difference analysis of the phase sequences of the two nodes. At each sampling time point, the difference between the phase values ​​of the two nodes is calculated, resulting in a phase difference sequence. If the two nodes maintain phase synchronization in this frequency band, the phase difference sequence exhibits stable characteristics, with the difference at each time point fluctuating within a small range. If there is a phase shift between the two nodes, the phase difference sequence exhibits fluctuating or drifting characteristics. By statistically analyzing the fluctuation amplitude of the phase difference sequence over the entire time axis, the degree of phase synchronization between the two nodes can be quantified; the smaller the fluctuation amplitude, the higher the synchronization degree, and the larger the fluctuation amplitude, the lower the synchronization degree.

[0086] In one possible implementation, the synchronization threshold is set based on the human ear's perception threshold for phase shift. The synchronization threshold is set to std(Δφ) < ​​5°, which is based on the human ear's binaural phase difference perception threshold (studies show that the lower limit of the phase difference that the human ear can perceive is approximately 3° to 6°, typically 5° as the judgment boundary). This value can be calibrated and adjusted within the range of 3° to 10° based on actual sound field test data. When the phase synchronization degree of a node pair is lower than this threshold, it indicates that the two nodes have produced a significant phase shift during the encoding process. This shift will affect the spatial positioning accuracy of the sound after signal reconstruction. Node pairs with phase synchronization degrees lower than the threshold are marked as phase-asynchronous node pairs, and the frequency band information shared by these node pairs is extracted to form a list of asynchronous frequency bands.

[0087] Preferably, the list of out-of-synchronization frequency bands records the frequency range, the node pair identifier, and the current phase synchronization value for each problematic frequency band. This list provides a clear target for subsequent frequency band migration, with each record in the list corresponding to a phase asynchronization problem to be addressed.

[0088] Understandably, the search for phase-coherent nodes involves finding alternative nodes with high phase synchronization with the problem frequency band within the current node allocation scheme. For each frequency band in the list of out-of-synchronization frequency bands, other nodes in the current allocation scheme are traversed, and the phase synchronization between that frequency band and each candidate node is calculated. If the phase synchronization between a candidate node and that frequency band exceeds a preset synchronization threshold, that node is marked as a phase-coherent node. Phase-coherent nodes have the ability to receive the migrated frequency band and can maintain a stable phase relationship with it. Furthermore, the frequency band migration process transfers the problem frequency band from the original bearer node to the phase-coherent node. The migration process includes updating the frequency band and node mapping relationship in the node allocation scheme, replacing the node identifier in the original record with the identifier of the phase-coherent node, updating the bearer frequency band list of the phase-coherent node, and adding the migrated frequency band to the node's processing task.

[0089] In one embodiment, after completing the migration operation of all asynchronous frequency bands, the mapping relationship between each frequency band and node in the node allocation scheme is updated to form an optimized node allocation scheme. In this optimized scheme, frequency bands that originally had phase asynchrony issues are reassigned to nodes whose phase characteristics match theirs, thereby improving the phase synchronization between nodes and laying the foundation for subsequent signal reconstruction and spatial sound field localization restoration.

[0090] Step S106: The frequency band signals of each node are distributedly encoded and recombined using an optimized node allocation scheme. The spatial sound field positioning of the recombined signal is extracted and compared with the positioning of the original signal to identify the final lossless audio signal that meets the preset positioning range.

[0091] For the reconstructed audio signal, the amplitude ratio between each channel is extracted as the sound level difference (ILD), and the arrival time difference between each channel is extracted as the time difference (ITD). Based on the ILD and ITD, the azimuth and elevation angles of the sound source in three-dimensional space are calculated. Specifically, the azimuth angle θ = arcsin((c × ITD) / d), where c is the speed of sound (343 m / s) and d is the head width (0.2 m); the elevation angle φ = atan(ILD / k), where k is an empirical constant of 1.5. This value is derived statistically from the slope of the linear relationship between elevation angle and sound level difference in multi-channel sound field calibration experiments and can be recalibrated according to the actual loudspeaker array configuration. The azimuth and elevation angles are combined to form the spatial sound field positioning of the reconstructed signal. The spatial sound field localization of the original multi-channel audio signal is extracted using the same method as the reconstructed signal, and used as the original localization reference. The spatial sound field localization of the reconstructed signal is then compared channel by channel with the original localization reference, and the difference between the two in azimuth and elevation angles is calculated to obtain the localization deviation. If the localization deviation is lower than a preset deviation threshold, the reconstructed audio signal is output as the final lossless audio signal; if the localization deviation exceeds the preset deviation threshold, the phase value of the channels in the reconstructed audio signal with excessive localization deviation is adjusted.

[0092] In distributed audio processing scenarios, the frequency band signals generated by each node after independent encoding are stored in the form of encoded bitstreams. The signal reconstruction process requires decoding and restoring these encoded bitstreams.

[0093] For example, the decoding operation is the inverse of the encoding operation. Entropy decoding is performed on the encoded bit stream to obtain quantized data, and then dequantization is performed on the quantized data to recover the sub-band signal amplitude values ​​of each frequency band.

[0094] Specifically, frequency domain splicing is the operation of concatenating the sub-band signals of each frequency band in the frequency domain in ascending order of frequency.

[0095] In one embodiment, the low-frequency sub-band signal occupies the low-frequency region of the spectrum, and the high-frequency sub-band signal occupies the high-frequency region of the spectrum. During splicing, the relative positions of each frequency band in the frequency domain remain unchanged to form a complete frequency domain signal representation. After completing the frequency domain splicing, an inverse Fourier transform is performed on the splicing result to restore the frequency domain signal to a reconstructed audio signal in the time domain.

[0096] It's important to note that spatial sound field localization relies on two key parameters: sound level difference and time difference. Sound level difference refers to the ratio of the amplitudes of the same sound source signal arriving at different channel speakers. When the sound source is biased to one side, the speaker closer to the source receives a larger amplitude signal, while the speaker farther away receives a smaller amplitude signal. This amplitude ratio reflects the degree of horizontal offset of the sound source. Time difference refers to the difference in the arrival times of the same sound source signal at different channel speakers. Due to the speed limitation of sound wave propagation, different distances from the sound source to each speaker result in different arrival times. This time difference reflects the distance relationship between the sound source and each speaker.

[0097] In one possible implementation, the azimuth angle is calculated based on a comprehensive analysis of sound level difference and time difference. Using the listener's position as the origin and the direction directly in front as the zero-degree reference direction, when the sound level difference indicates a larger amplitude in the left channel and the time difference indicates the left channel signal arrives first, the azimuth angle of the sound source is located in the left region; when the sound level difference indicates a larger amplitude in the right channel and the time difference indicates the right channel signal arrives first, the azimuth angle of the sound source is located in the right region. The elevation angle is calculated based on the relationship between the sound level difference and time difference between the upper and lower channels, using similar logic to determine the angular position of the sound source in the vertical direction.

[0098] Preferably, the extraction of the original positioning reference uses the same sound level difference and time difference calculation method as the reconstructed signal. Before performing distributed processing on the original multi-channel audio signal, the sound level difference and time difference between each channel are extracted, and the azimuth and elevation angles of the original signal are calculated as the positioning reference. This positioning reference records the true spatial location information of the sound source in the original signal, serving as a reference standard for evaluating the positioning accuracy of the reconstructed signal.

[0099] Understandably, the positioning comparison process involves calculating the difference between the spatial sound field positioning of the reconstructed signal and the original positioning reference channel by channel and dimension by dimension. For each channel, the difference between the azimuth angle and the original signal, as well as the difference between the elevation angle and the original signal, are calculated. These angular differences from each channel are then summed to form the positioning deviation. The smaller the positioning deviation, the more accurate the reconstruction of the original signal's spatial location. Furthermore, the deviation threshold is set based on the human ear's ability to perceive and distinguish spatial positioning. When the positioning deviation is below the perceptible angular threshold, listeners cannot detect the difference in spatial positioning between the reconstructed and original signals, and the reconstructed audio signal meets the requirements for lossless reconstruction. When the positioning deviation exceeds this threshold, listeners will perceive a shift in the sound source's location, requiring correction processing of the reconstructed signal.

[0100] In one embodiment, phase compensation correction corrects positioning deviations by adjusting the phase value of a specific channel in the reconstructed signal. The adjustment amount is determined based on the magnitude and direction of the positioning deviation. When the positioning of a channel is biased to the left, the phase of that channel is appropriately delayed to reduce the left bias; when the positioning of a channel is biased to the right, the phase of that channel is appropriately advanced to reduce the right bias. After phase adjustment, the spatial sound field positioning of the adjusted signal is re-extracted and compared with the original positioning reference. If the positioning deviation still exceeds the deviation threshold, adjustment continues until the positioning deviation meets the requirements, at which point the final lossless audio signal is output. A preset deviation threshold T is used. loc The typical value is 3° (based on the fact that the minimum discernible angle for human hearing to perceive the direction of a sound source is approximately 1° to 5°, 3° is taken as the tolerance boundary for engineering implementation, which can be adjusted according to the accuracy of the target sound field reconstruction). If the positioning deviation still exceeds T loc Then continue to adjust, with the maximum number of iterations set to 10 (typical value, which can be adjusted according to real-time processing delay requirements); if the positioning deviation still does not meet the requirements after the number of iterations reaches the upper limit, the current optimal compensation result is output and a warning mark is recorded, indicating that the positioning deviation of the channel exceeds the compensable range, and the upper-level system decides whether to re-execute the encoding process.

[0101] Secondly, embodiments of the present invention also include a lossless encoding and decoding device for distributed audio and video processing, the device comprising:

[0102] The memory, the processor, and the computer program stored on the memory and executable on the processor, the computer program being configured to implement the lossless encoding and decoding steps of a distributed audio and video processing as described above.

[0103] Thirdly, embodiments of the present invention also disclose a medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the aforementioned lossless encoding and decoding method for distributed audio and video processing.

[0104] If the technical solution of this application involves the collection, processing, or application of personal information, the relevant products have, before implementing any personal information processing activities, fully and clearly informed individuals of the processing rules in accordance with the "Personal Information Protection Law of the People's Republic of China" and other current laws and regulations, and obtained their voluntary and explicit consent. If sensitive personal information is involved, the product has obtained the individual's separate consent before processing, and such consent is given in an explicit manner. For example, prominent signs are set up in the area where information collection devices such as cameras are located, clearly indicating "Entering is considered as consent to the collection of personal information"; or through pop-ups, checkboxes, user-initiated uploads, etc., under the premise of clearly listing the processor's identity, processing purpose, processing method, and information type, the user actively completes the authorization operation. The above mechanisms ensure that all personal information processing activities are based on legal authorization and fully comply with national compliance requirements regarding personal information protection.

[0105] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

Claims

1. A lossless encoding and decoding method for distributed audio and video processing, characterized in that, include: Obtain the spatial coordinates of the physical loudspeakers in each channel, construct a channel topology diagram, extract the coherence level between channels based on the channel topology diagram, mark the frequency bands with coherence levels below a preset threshold as phase-sensitive frequency bands, and generate a combination of phase-sensitive frequency bands. Based on the combination of phase-sensitive frequency bands, the phase-locking state of the low-frequency and high-frequency components of the same channel is analyzed, and a preliminary node allocation scheme is generated based on the frequency band grouping boundary. Based on the node allocation scheme, the multi-channel audio is split into frequency bands, and cross-node phase synchronization detection is performed on the frequency band signals of each node; An optimized node allocation scheme is used to perform distributed coding and reassembly on the frequency band signals of each node.

2. The method according to claim 1, characterized in that, The extraction of inter-channel coherence levels, marking frequency bands with coherence levels below a preset threshold as phase-sensitive frequency bands, and generating a combination of phase-sensitive frequency bands specifically includes: Traverse each pair of adjacent channel nodes in the channel topology diagram, extract the audio signals of adjacent channels within the same time window, perform a short-time Fourier transform on the audio signals to obtain the amplitude and phase of each frequency component, multiply the frequency components of two adjacent channels point by point for each frequency band, sum and take the average value to obtain the coherence value, if the coherence value is lower than the preset coherence threshold, then mark the frequency band as a phase-sensitive frequency band; The number of channel pairs marked as phase-sensitive frequency bands in the same frequency band in the channel topology diagram is counted, and the frequency bands whose number of channel pairs exceeds a preset threshold are classified into phase-sensitive frequency band combinations. The phase-sensitive frequency band combination includes the frequency range of each sensitive frequency band and its corresponding channel pair association information.

3. The method according to claim 1, characterized in that, The step of analyzing the phase-locking state between the low-frequency and high-frequency components of the same channel based on the combination of phase-sensitive frequency bands specifically includes: The audio signal within the same channel is extracted based on the phase-sensitive frequency band combination. The audio signal is then subjected to a short-time Fourier transform to obtain a time-frequency representation. The signal is divided into low-frequency components and high-frequency components according to their frequency. The instantaneous phase values ​​of the low-frequency components and high-frequency components at the same time point are extracted to obtain a low-frequency phase sequence and a high-frequency phase sequence. The phase difference between the low-frequency phase sequence and the high-frequency phase sequence at the same time point is calculated. The change amplitude of the phase difference over a continuous time period is statistically analyzed. If the change amplitude is lower than a preset stability threshold, the low-frequency component and the high-frequency component in the corresponding time period are determined to be in a phase-locked state. If the change amplitude exceeds the preset stability threshold, the time period is determined to be in a phase-unlocked state. The frequency position corresponding to the phase-unlocked state is recorded as the unlocked position.

4. The method according to claim 3, characterized in that, Also includes: Based on the distribution of the unlocked positions on the frequency axis, the unlocked positions are used as candidate points for the frequency band group boundary. For each sensitive frequency band combination, the number of unlocked positions of its low-frequency components and the number of unlocked positions of its high-frequency components are counted. The unlocked ratio is obtained by dividing the number of unlocked positions of the low-frequency components by the number of unlocked positions of the high-frequency components. If the unlock ratio exceeds the preset allowable threshold, the sensitive frequency band combination is merged with the adjacent sensitive frequency band combination on the frequency axis. The unlock position is re-detected and the boundary is delineated within the merged frequency band range to obtain the frequency band group boundary.

5. The method according to claim 1, characterized in that, The process of generating a preliminary node allocation scheme based on frequency band grouping boundaries specifically includes: Based on the frequency band grouping boundary, obtain the frequency range and bandwidth information of each frequency band, extract the list of available nodes from the distributed processing cluster, establish a frequency band node mapping table according to the correspondence between the number of frequency bands and the number of nodes, the frequency band node mapping table records the target node identifier and the bandwidth value of each frequency band, and sequentially allocate each frequency band within the frequency band grouping boundary to each node in the node list to obtain a preliminary node allocation scheme; According to the preliminary node allocation scheme, each node is traversed, and the bandwidth value of the frequency band carried by each node is accumulated as the processing load of that node. The processing load is compared with a preset high load threshold. If the processing load exceeds the preset high load threshold, the node is marked as a high load node. If the processing load is lower than the preset low load threshold, the node is marked as a low load node. For the frequency bands carried by the high-load nodes, the frequency bands to be migrated are extracted one by one in descending order of the bandwidth values ​​recorded in the frequency band node mapping table. The node with the smallest current processing load is selected from the low-load nodes as the receiving node. The frequency bands to be migrated are migrated from the high-load nodes to the receiving nodes, and the node identifiers in the frequency band node mapping table are updated to obtain the node allocation scheme.

6. The method according to claim 1, characterized in that, The frequency band splitting of multi-channel audio based on the node allocation scheme specifically includes: According to the node allocation scheme, obtain the frequency band list carried by each node, perform bandpass filtering on each channel signal of the multi-channel audio according to the frequency range recorded in the frequency band list, extract the sub-band signal of the corresponding frequency band from each channel signal, and distribute the sub-band signal to the corresponding node according to the node identifier to obtain the sub-band signal to be encoded for each node. For the sub-band signal to be encoded at each node, each node is driven to perform independent encoding. The independent encoding performs quantization and entropy encoding on the sub-band signal. After each node completes independent encoding, it outputs the corresponding encoded bit stream as the frequency band signal of each node, thus obtaining the frequency band signal of each node.

7. The method according to claim 1, characterized in that, The cross-node phase synchronization detection of signals in each node frequency band specifically includes: The phase information of the frequency band signal carried by each node is obtained. For every two adjacent nodes, a node pair is constructed. For each frequency band, the phase sequence of the two nodes in the node pair in that frequency band is extracted. The phase difference is calculated point by point in time. The standard deviation of the phase difference is calculated. If the standard deviation of the phase difference is lower than the preset synchronization threshold, the node pair is phase synchronized. Otherwise, it is marked as a phase-asynchronous node pair. The frequency band identifier shared by the two nodes in the phase-asynchronous node pair is extracted. The frequency band identifier is associated with the corresponding node pair information and recorded to obtain a list of asynchronous frequency bands. According to the list of asynchronous frequency bands, each pair of asynchronous nodes is traversed. For each asynchronous frequency band, the phase sequence of other nodes in the current node allocation scheme is retrieved from the list. The standard deviation between the phase sequence of the relevant node in the frequency band and the original phase sequence of the frequency band is calculated. If the standard deviation is lower than the preset synchronization threshold, it is marked as a phase coherent node. The asynchronous frequency bands are migrated from the original bearer nodes to the phase coherent nodes, and the frequency band and node mapping relationship in the node allocation scheme is updated to obtain an optimized node allocation scheme.

8. The method according to claim 1, characterized in that, The process of performing distributed coding and reassembly on the frequency band signals of each node using an optimized node allocation scheme specifically includes: For the reconstructed audio signal, the amplitude ratio between each channel is extracted as the sound level difference, and the arrival time difference between each channel is extracted as the time difference. Based on the sound level difference and the time difference, the azimuth and elevation angles of the sound source in three-dimensional space are calculated. The azimuth angle is calculated as the sound speed multiplied by the time difference divided by the arcsine of the head width. The elevation angle is calculated as the sound level difference divided by the arctangent of an empirical constant. The azimuth and elevation angles are combined to form the spatial sound field positioning of the reconstructed signal. The spatial sound field positioning of the original multi-channel audio signal is extracted in the same way as the reconstructed signal and used as the original positioning reference. The spatial sound field positioning of the reconstructed signal is compared with the original positioning reference channel by channel, and the difference between the two in azimuth and elevation angles is calculated to obtain the positioning deviation.

9. A lossless encoding and decoding device for distributed audio and video processing, characterized in that, The device includes: A memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of lossless encoding and decoding of a distributed audio and video processing as claimed in any one of claims 1 to 8.

10. A medium, characterized in that, The medium is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of a lossless encoding and decoding method for distributed audio and video processing as described in any one of claims 1 to 8.

Images