A method for multipath transmission of audio data between LE-Audio devices

By establishing multi-path redundant transmission and fragment-level hash tree verification between LE-Audio devices, combined with interleaved connection isochronous streams, the problems of data packet loss and integrity verification in LE-Audio audio data transmission are solved, achieving efficient audio data transmission and reliability assurance.

CN123124132APending Publication Date: 2026-11-17MINAMI ACOUSTICS LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611198457.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-07
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing LE-Audio audio data transmission schemes have shortcomings in multi-path redundant transmission, real-time integrity verification at the fragment level, and verifiability of transmission records, leading to data packet loss or transmission errors, which affect the integrity and real-time performance of audio playback.

Method used

By establishing multi-path redundant transmission and fragment-level hash tree verification between LE-Audio devices, utilizing lightweight neural networks to optimize encoding, constructing hierarchical hash trees and generating transmission metadata, and combining interleaved connection isochronous streams for audio data transmission, real-time integrity verification and tamper-proof recording of data fragments are achieved.

Benefits of technology

It improves the anti-interference capability and reliability of audio transmission, reduces verification latency and buffering overhead, and ensures the integrity of audio data and the traceability of the transmission process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123124132A_ABST
    Figure CN123124132A_ABST
Patent Text Reader

Abstract

This invention relates to a method and apparatus for multipath transmission of audio data between LE-Audio devices, belonging to the field of Bluetooth Low Energy audio technology. The invention establishes a connection isochronous group containing at least two connection isochronous streams between a first LE-Audio device and a second LE-Audio device, forming multiple parallel transmission paths. After splitting the audio data packet into N data fragments, a hierarchical hash tree is constructed using the hash value of each fragment as a leaf node, obtaining the root hash value and generating transmission metadata. The same data fragment is redundantly transmitted via at least two different connection isochronous streams. The receiving end can use the fragment hash value and authentication path to verify the integrity of the received data fragments piece by piece. Through the synergy of multipath redundant transmission and fragment-level hash tree verification, the reliability of audio transmission and the efficiency of data integrity verification are improved while maintaining the low-power characteristics of LE-Audio.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of Bluetooth Low Energy audio technology, and in particular to a method for multipath transmission of audio data between LE-Audio devices. Background Technology

[0002] In the Bluetooth Low Energy Audio (LE-Audio) technology standard, the Bluetooth core specification introduces transmission mechanisms such as Connected Isochronous Stream (CIS) and Connected Isochronous Group (CIG), supporting the transmission of multiple synchronous audio streams between an audio source device and one or more audio receiving devices. It also supports multi-connection data stream transmission with one master and multiple slaves, and each connected synchronization group can include multiple connected synchronization streams. Compared to the traditional classic Bluetooth Audio (BR / EDR-Audio) solution, LE-Audio has achieved significant progress in synchronous transmission and power consumption control.

[0003] In real-world applications, LE-Audio devices often face challenges such as channel quality fluctuations and sudden interference, leading to packet loss or transmission errors on a single CIS link, thus affecting the integrity and real-time performance of audio playback. To address these issues, several improvements exist in existing technologies. For example, some solutions propose a multi-channel transmission architecture based on different access addresses within the Bluetooth system, resolving channel conflicts and power consumption issues in multi-channel transmission through channel allocation. Other solutions leverage the characteristics of CIS links and Broadcast Isochronous Stream (BIS) links in LE-Audio to extend the topology of the Bluetooth Low Energy audio network, thereby improving communication reliability in multi-point to multi-point bidirectional communication scenarios.

[0004] Regarding audio data integrity verification, existing solutions generally employ methods such as calculating checksums for complete audio frames or performing XOR checks on Bluetooth packets. While these methods can verify the correctness of audio data to some extent, they typically require overall verification of the entire data packet. When some data fragments arrive before others in a multi-path scenario, the receiving end cannot immediately verify the integrity of the arrived fragments and must wait until all fragments have been received before verification can be performed, resulting in additional latency and buffering overhead. Furthermore, existing LE-Audio transmission solutions lack corresponding measures for preventing tampering during audio data transmission and ensuring post-transmission verifiability.

[0005] In summary, existing LE-Audio audio data transmission schemes still have significant room for improvement in terms of multi-path redundant transmission, real-time integrity verification at the fragment level, and verifiability of transmission records. Summary of the Invention

[0006] The main objective of this invention is to provide a method for multipath transmission of audio data between LE-Audio devices. By combining multipath redundant transmission with fragment-level hash tree verification, the reliability of audio transmission and the efficiency of data integrity verification are improved while maintaining the low power consumption characteristics of LE-Audio.

[0007] To achieve the above objectives, the present invention provides a method for multipath transmission of audio data between LE-Audio devices, comprising the following steps: The first LE-Audio device and the second LE-Audio device establish a connection isochronous group based on the Bluetooth Low Energy Audio protocol. The connection isochronous group is configured to include at least two independent connection isochronous streams, forming multiple parallel audio data transmission paths, and each connection isochronous stream has at least one sub-event in each isochronous interval. The first LE-Audio device performs encoding processing on the raw audio data to generate audio data packets conforming to the BLE-Audio specification, and splits the audio data packets into N data fragments, where N is an integer greater than or equal to 2; The first LE-Audio device calculates the hash value of each data segment, uses the hash values ​​of the N data segments as leaf nodes, constructs a hierarchical hash tree by calculating the hash values ​​of the parent nodes layer by layer, obtains the root hash value of the hierarchical hash tree, and generates the transmission metadata of the audio data packet. The transmission metadata includes at least the audio packet sequence number and the number of segments N. The first LE-Audio device sends the data fragments through the at least two isochronous streams, wherein the same data fragment is transmitted via at least two different isochronous streams. The first LE-Audio device associates the root hash value with the transmission metadata and writes it into the local anti-tampering log to generate a corresponding verification record identifier.

[0008] Furthermore, the step of the first LE-Audio device performing encoding processing on the raw audio data to generate audio data packets conforming to the BLE-Audio specification includes: The first LE-Audio device acquires raw audio data and extracts acoustic features from the raw audio data, the acoustic features including spectral flux F. t and the spectral centroid C t ; Where the spectral flux F t Satisfy: F t =Σ(W k ·|X(t,k)|), k=1 to M, W k X(t,k) is the weighting coefficient of the k-th frequency band, X(t,k) is the spectral amplitude of the k-th frequency band in the t-th frame, and M is the total number of frequency bands; The acoustic features are input into a preset lightweight neural network, which then outputs the importance coefficient α of each audio frame in the original audio data. t The importance coefficient α t Satisfying 0≤α t ≤1, the lightweight neural network is configured as an end-to-end structure with the acoustic features as input and the importance coefficient as output; Based on the importance coefficient α t Assign coding bits to each audio frame, and assign B coding bits to the t-th frame. t =B base +α t ×(B max -B base ), where B base Based on the number of basic encoding bits, B max To maximize the number of encoded bits, the original audio data is encoded to generate the audio data packet conforming to the BLE-Audio specification.

[0009] Furthermore, the step of inputting the acoustic features into a preset lightweight neural network, and having the lightweight neural network output the importance coefficient αt of each audio frame in the original audio data, includes: The lightweight neural network includes an input layer, at least one convolutional layer, at least one recurrent layer, and an output layer, wherein the convolutional layer is configured to extract local features from the acoustic features, and the recurrent layer is configured to extract temporal context information from the feature sequence output by the convolutional layer. The acoustic features of frame t are input into the lightweight neural network, and the lightweight neural network outputs the original importance score r corresponding to that frame. t Original importance score r t For unnormalized real numbers; According to the r t The importance coefficient α is calculated using the BatchNorm normalization function, along with the total number of audio frames T. t , satisfying: α t =Sigmoid((r t -μ r ) / σ r ) Where μ r For the r t The mean, σ r For the r t The standard deviation of α, where sigmoid is the preset activation function, such that α t Satisfying 0≤α t ≤1.

[0010] Furthermore, the step of constructing a hierarchical hash tree by calculating the hash values ​​of parent nodes layer by layer to obtain the root hash value of the hierarchical hash tree includes: The N data fragments are used as the N leaf nodes of the 0th layer, and the hash value H(0,i) of each leaf node in the 0th layer is Hash(D). i ), where D i This represents the i-th data fragment, where 0 ≤ i ≤ N-1, and Hash represents the preset hash function; For a node at level k, the hash value of the parent node at level k+1 is calculated using the hash values ​​of two adjacent nodes as input, satisfying H(k+1,j)=Hash(H(k,2j)‖H(k,2j+1)), where k≥0 and j≥0; When only a unique parent node hash value is calculated for the (k+1)th layer, the unique parent node hash value is used as the root hash value of the hierarchical hash tree.

[0011] Furthermore, the step of the first LE-Audio device sending the data fragments through the at least two isochronous streams includes: The first LE-Audio device allocates the N data fragments to the at least two connection isochronous streams, and the same data fragment is allocated to at least two different connection isochronous streams; At least one sub-event sends the allocated data fragment within the isochronous interval of each connection isochronous stream, wherein the isochronous intervals of different connection isochronous streams are staggered on the time axis; The transmission mode of the connected isochronous streams is configured as an interleaved mode, in which the sub-events of the at least two connected isochronous streams are sent alternately in a polling manner.

[0012] Furthermore, the step of the first LE-Audio device associating the root hash value with the transmission metadata and writing it into the local anti-tampering log to generate a corresponding verification record identifier includes: The first LE-Audio device combines the root hash value corresponding to the current audio data packet, the transmission metadata, and the previous verification record identifier into the current verification record; A preset hash function is performed on the current verification record to obtain the current verification record identifier, which satisfies: Rid(i) = Hash(Root(i) || Meta(i) || Rid(i-1)); Where Rid(i) is the verification record identifier corresponding to the current audio data packet, Root(i) is the root hash value corresponding to the current audio data packet, Meta(i) is the transmission metadata corresponding to the current audio data packet, Rid(i-1) is the previous verification record identifier, || represents the data connection, and Hash is the preset hash function; The current verification record and its identifier are written into the local anti-tampering log. The local anti-tampering log is a linear data structure that only appends to writes. The second LE-Audio device verifies the integrity of the received data fragments by obtaining the verification record identifier and the corresponding verification record in the local anti-tampering log.

[0013] Furthermore, the step of configuring the connected isochronous group to include at least two independent connected isochronous streams to form multiple parallel audio data transmission paths includes: The first LE-Audio device and the second LE-Audio device establish the connection isochronous group based on the Bluetooth Low Energy Audio protocol. The connection isochronous group includes a first connection isochronous stream and a second connection isochronous stream. The first connection isochronous stream includes a first access address and a first channel resource, and the second connection isochronous stream includes a second access address and a second channel resource. The first connection isochronous stream and the second connection isochronous stream share the same isochronous interval parameter. Within the same time interval, there is a preset offset between the anchor time of the first connected isochronous stream and the anchor time of the second connected isochronous stream, and the first channel resource and the second channel resource correspond to different physical channels respectively; The sub-events of the first connection isochronous stream and the sub-events of the second connection isochronous stream are arranged alternately on the time axis. The transmission modes of the first connection isochronous stream and the second connection isochronous stream are both configured as interleaved, forming the multiple audio data transmission paths that are logically parallel.

[0014] The present invention also provides an inter-LE-Audio audio data multipath transmission apparatus, which performs the above-described inter-LE-Audio audio data multipath transmission method, wherein the inter-LE-Audio audio data multipath transmission apparatus comprises: A connection establishment unit is used to establish an isochronous group with a second LE-Audio device based on the Bluetooth Low Energy Audio protocol; The encoding fragmentation unit is used to perform encoding processing on the original audio data, generate audio data packets conforming to the BLE-Audio specification, and split the audio data packets into N data fragments, where N is an integer greater than or equal to 2; The hash tree construction unit is used to calculate the hash value of each data fragment, with the hash values ​​of the N data fragments as leaf nodes, and to construct a hierarchical hash tree by calculating the hash values ​​of the parent nodes layer by layer, thereby obtaining the root hash value of the hierarchical hash tree and generating the transmission metadata of the audio data packet. A multipath transmission unit is used to transmit the data fragments respectively through the at least two connection isochronous streams, wherein the same data fragment is transmitted via at least two different connection isochronous streams; The log writing unit is used to associate the root hash value with the transmission metadata and write it into the local anti-tampering log to generate a corresponding verification record identifier.

[0015] The present invention 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 steps of the above-described LE-Audio device-to-device audio data multipath transmission method.

[0016] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for multipath transmission of audio data between LE-Audio devices.

[0017] The multipath transmission method for audio data between LE-Audio devices provided in the above embodiments has the following beneficial effects: By establishing at least two isochronous streams within the same isochronous connection group to form multiple parallel transmission paths, the same data fragments are redundantly sent through different paths. When a single link is interfered with and loses packets, the fragments of other paths can compensate for the loss, effectively improving the anti-interference capability and reliability of audio transmission. By constructing a hierarchical hash tree using the hash values ​​of each data fragment obtained from splitting audio data packets as leaf nodes, the root hash value is transmitted along with the data. The receiving end can use the fragment hash value of a single fragment and the authentication path to verify the integrity of the fragment piece by piece without waiting for all fragments to be received, thus reducing verification latency and buffering overhead. By associating the root hash value with the transmission metadata and writing it into the local anti-tampering log, a chain of verification record identifiers is generated, making the transmission record traceable and tamper-proof, which facilitates the receiving end to perform post-transmission verification of the integrity of the audio data packet transmission process. Attached Figure Description

[0018] Figure 1 This is a flowchart of a method for multipath transmission of audio data between LE-Audio devices according to an embodiment of the present invention; Figure 2 This is a structural block diagram of an LE-Audio device for multipath transmission of audio data between devices according to an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] Reference Figure 1 This is a flowchart illustrating a method for multipath transmission of audio data between LE-Audio devices proposed in this invention, comprising the following steps: S1, the first LE-Audio device and the second LE-Audio device establish a connection isochronous group based on the Bluetooth Low Energy Audio protocol. The connection isochronous group is configured to include at least two independent connection isochronous streams, forming multiple parallel audio data transmission paths, and each connection isochronous stream has at least one sub-event in each isochronous interval. S2, the first LE-Audio device performs encoding processing on the raw audio data to generate an audio data packet conforming to the BLE-Audio specification, and splits the audio data packet into N data fragments, where N is an integer greater than or equal to 2; S3, the first LE-Audio device calculates the hash value of each data segment, uses the hash values ​​of the N data segments as leaf nodes, constructs a hierarchical hash tree by calculating the hash values ​​of the parent nodes layer by layer, obtains the root hash value of the hierarchical hash tree, and generates the transmission metadata of the audio data packet. The transmission metadata includes at least the audio packet sequence number and the number of segments N. S4, the first LE-Audio device sends the data fragments through the at least two isochronous streams of the connection respectively, wherein the same data fragment is transmitted via at least two different isochronous streams of the connection; S5, the first LE-Audio device associates the root hash value with the transmission metadata and writes it into the local anti-tampering log to generate a corresponding verification record identifier.

[0021] As described in step S1 above, the first LE-Audio device and the second LE-Audio device establish a connection isochronous group via Bluetooth Low Energy Audio protocol. This connection isochronous group is configured at the link layer to carry at least two independent connection isochronous streams. Each connection isochronous stream corresponds to a logical audio data transmission path, and multiple connection isochronous streams together constitute a set of parallel transmission paths. During the connection establishment phase, the first LE-Audio device, as the master device, initiates a connection request to the second LE-Audio device. After both parties complete the synchronization parameter negotiation, they determine the isochronous interval parameters, the number of sub-events, and the transmission mode of the connection isochronous group. Each connection isochronous stream is assigned an independent access address, and each connection isochronous stream has its own independent anchor point time within the same isochronous interval. Different anchor points are staggered by a preset offset, so that the sub-events of each connection isochronous stream are arranged in an interleaved manner on the timeline. After negotiation and configuration, the transmission mode of both connection isochronous streams is set to interleaved mode. In this mode, multiple connection isochronous streams within the same connection isochronous group alternately occupy air interface resources at the sub-event granularity. After a connection isochronous stream completes the transmission of a sub-event, it immediately switches to the sub-event of the next connection isochronous stream. At the same time, the physical channels corresponding to different connection isochronous streams are allocated to different frequency hopping sequences or different channel frequencies to reduce co-channel interference between links. In this way, a parallel transmission architecture consisting of multiple connection isochronous streams is established between the first LE-Audio device and the second LE-Audio device. Each connection isochronous stream contains at least one sub-event available for data transmission within each isochronous interval.

[0022] As described in step S2 above, after the first LE-Audio device acquires the raw audio data, it calls the encoder to perform compression encoding on the audio stream according to the Bluetooth Low Energy Audio specification. The encoder can use a low-complexity communication codec or other compatible audio codec formats to output audio data packets conforming to the BLE-Audio packet format. After encoding is completed, the first LE-Audio device divides the data into several continuous data segments along the logical boundaries of the audio packets. The division can be based on a fixed byte length or alignment according to the natural boundaries of the audio frames. The total number of segments N is determined by the actual length of the audio data packets and the preset segmentation granularity. In addition to carrying the corresponding audio payload segment, each data fragment also includes fragment header information. The header information includes the packet sequence number of the audio data packet to which the fragment belongs, the position number of the fragment within the packet, and the total number of fragments N. The addition of fragment header information allows the receiving end to know the exact position of the fragment in the original audio data packet even when only a portion of the fragment is received. The payload segments of all data fragments can be completely restored to the original payload of an audio data packet after being spliced ​​in sequence. The fragmentation operation does not change the bit structure of the underlying encoded data; it only performs encapsulation and segmentation at the outer layer.

[0023] As described in step S3 above, the first LE-Audio device performs hash operations on each of the N data fragments obtained in step S2, generating a fixed-length hash value for each data fragment as its digital digest. The hash operation can use a lightweight hash function with low computational overhead to adapt to the resource constraints of the embedded device. After all N fragment hash values ​​have been calculated, these N hash values ​​are considered the leaf nodes at the bottom level of the hierarchical hash tree structure. Starting from the level where the leaf node is located, the hash values ​​are merged layer by layer upwards. The hash values ​​of two adjacent nodes in the same level are concatenated and input into the same hash function again to obtain the hash value of the parent node of the next higher level. If the number of nodes in the current level is odd, the last node is paired with its own copy value to participate in the calculation. The layer-by-layer merging process continues, with the number of nodes halved each time the layer moves up, until only one node remains in a certain level. The hash value of this node is the root hash value of the entire hierarchical hash tree. The binary length of the root hash value is consistent with the hash value of a single fragment and does not change with the increase of the number of fragments N. After the hash tree is constructed, the first LE-Audio device combines the root hash value with the key context information of the current audio data packet to generate transmission metadata. This metadata includes the audio packet sequence number to identify the transmission order of the audio data packets, the number of fragments N to inform the receiving end of the leaf node size of the hash tree, and may also include encoding type identifiers and time reference information. The transmission metadata is encapsulated into an independent metadata message, which is sent along with the data fragments to the second LE-Audio device, providing a baseline check value for the receiving end's fragment-by-fragment integrity verification.

[0024] As described in step S4 above, the first LE-Audio device distributes the N data fragments obtained in step S2 to multiple isochronous streams established in step S1 for transmission. Each data fragment is simultaneously arranged into the transmission queues of at least two isochronous streams. Within the equal time interval of each isochronous stream, the sending scheduler sequentially loads data fragments from the queue into link layer data packets according to the time window of the sub-event and pushes them to the air interface. Since there is a preset offset between the anchor points of different isochronous streams and the transmission mode is interleaved, the sub-events of multiple isochronous streams appear alternately on the time axis. When the same data fragment is sent through different isochronous streams, it is naturally dispersed in time sequence. When a data fragment is lost due to a sudden interference in the time slot of a sub-event on one isochronous stream, the corresponding redundant fragment on another isochronous stream is still transmitted independently in subsequent sub-events. Each data fragment retains its original fragment position number and the sequence number of the audio packet it belongs to when it is encapsulated by the link layer. The copies of the same data fragment carried on different isochronous streams carry completely identical sequence identifiers. The receiving end can use this to associate and deduplicate the same fragment copies from different paths. Multiple connections share the same isochronous interval parameter, and the transmission cycles of each path remain synchronized. The transmission process of fragments on each path is executed at the same isochronous interval level, without introducing additional latency overhead. In this way, N data fragments are transmitted independently in a redundant manner on multiple parallel connections with isochronous intervals, forming a fragment-level multi-path redundant distribution at the air interface level.

[0025] As described in step S5 above, the first LE-Audio device combines and binds the root hash value calculated in step S3 with the transmission metadata. The bound structure is written as a verification record entry into the device's locally maintained anti-tampering log. The write operation is performed before or simultaneously with the data fragment transmission, without blocking the audio data packet transmission process. The local anti-tampering log adopts an append-only linear storage structure. Each verification record entry is sequentially appended to the log. The written records cannot be deleted or overwritten, and the log's own storage area is protected by hardware-level access. In addition to the root hash value and transmission metadata, each verification record entry also carries a pointer to the previous verification record identifier at the entry header, forming a chain of verification records. The content of the current verification record entry is hashed to generate a fixed-length verification record identifier. This verification record identifier serves as both the retrieval index for the current record and the link target pointed to by the pointer in the next verification record entry. The verification record identifier is sent to the second LE-Audio device along with the data fragments via the connection isochronous stream, or transmitted separately through an independent control sub-event. After the second LE-Audio device receiver obtains the verification record identifier, it can read the corresponding verification record entry from the local anti-tampering log based on the identifier, extract the root hash value stored therein, and compare it with the root value of the hash tree reconstructed by the receiver itself. This verifies whether the received data fragments remain intact after multi-path transmission. The time order of each verification record in the log is consistent with the sending order of the audio data packets, providing a traceable integrity audit clue for the transmission process.

[0026] In one embodiment, the first LE-Audio device and the second LE-Audio device establish a connection isochronous group based on the Bluetooth Low Energy Audio protocol. The connection isochronous group is configured to include at least two independent connection isochronous streams, forming multiple parallel audio data transmission paths, and each connection isochronous stream has at least one sub-event in each isochronous interval. Step S1 includes: S11, Establish a connection isochronous group and negotiate isochronous interval parameters; S12, assigns independent access addresses and anchor point offsets to each connection's isochronous stream; S13, configure the transmission mode of each connection's isochronous stream to interleaved mode; S14 allocates different physical channel resources to each connection's isochronous stream.

[0027] In the specific implementation process, the first LE-Audio device, acting as the master device, sends a connection isochronous group request message to the second LE-Audio device, which is acting as the slave device, based on the established asynchronous connectionless link. The message carries a set of isochronous interval parameters supported by the master device. Each parameter corresponds to a different audio transmission delay and bandwidth configuration. The second LE-Audio device selects a matching isochronous interval parameter from the parameter set according to its own hardware processing capabilities and audio playback requirements, and returns it to the first LE-Audio device through a connection isochronous group response message.

[0028] After receiving the response message, the first LE-Audio device generates an independent access address for each connection isochronous stream. The access address is a 32-bit binary value, following the access address generation rules specified in the Bluetooth core specification. The first LE-Audio device uses a pseudo-random number generator to generate the high 24 bits of the access address, and sets the low 8 bits to a fixed value corresponding to the connection isochronous stream index. The low 8 bits of the access address for the first connection isochronous stream are set to 0x01, and the low 8 bits of the access address for the second connection isochronous stream are set to 0x02. The generated access address is sent to the second LE-Audio device through a connection isochronous group configuration message. The second LE-Audio device stores the access address in its local link layer register for synchronization and identification during data reception.

[0029] The first LE-Audio device calculates the anchor offset of each isochronous stream based on the number of connected isochronous streams and the duration of sub-events in each connected isochronous stream. The anchor offset refers to the time difference between the anchor time of each connected isochronous stream and the starting anchor time of the connected isochronous group. For a connected isochronous group containing two connected isochronous streams, the first LE-Audio device sets the anchor offset of the first connected isochronous stream to 0 microseconds and the anchor offset of the second connected isochronous stream to half the duration of a single sub-event. The duration of a single sub-event is determined by the isochronous interval parameter and the number of sub-events in each isochronous interval. When the isochronous interval is 10 milliseconds and each isochronous interval contains 2 sub-events, the duration of a single sub-event is 5 milliseconds, and the anchor offset of the second connected isochronous stream is set to 2500 microseconds. The first LE-Audio device sends the calculated anchor offset to the second LE-Audio device through a configuration message. The second LE-Audio device adjusts its local clock according to the anchor offset to achieve multi-stream time synchronization with the first LE-Audio device.

[0030] The first LE-Audio device configures the transmission mode of all connection isochronous streams to interleaved mode. In interleaved mode, multiple connection isochronous streams within the same connection isochronous group alternately occupy air interface resources at the sub-event granularity. The link layer scheduler of the first LE-Audio device schedules the sub-events of each connection isochronous stream sequentially according to the index order of the connection isochronous streams. When a connection isochronous stream completes the transmission of a sub-event, the scheduler immediately switches to the sub-event of the next connection isochronous stream. The link layer of the second LE-Audio device opens a receive window within the corresponding time window according to the pre-configured anchor offset and sub-event duration to receive data from the corresponding connection isochronous stream.

[0031] The first LE-Audio device allocates independent physical channel resources for each connected isochronous stream. The physical channel resources include frequency hopping sequences and channel mapping tables. The first LE-Audio device generates two different frequency hopping sequences, which are assigned to the first connected isochronous stream and the second connected isochronous stream, respectively. The frequency hopping sequences are generated based on the frequency hopping algorithm specified in the Bluetooth core specification, using different initialization parameters. The first LE-Audio device also generates an independent channel mapping table for each connected isochronous stream. The channel mapping table marks the available physical channels and the disabled physical channels. The channel mapping tables of the two connected isochronous streams can partially overlap. The first LE-Audio device ensures that the two connected isochronous streams will not use the same physical channel for transmission at the same time. The first LE-Audio device sends the frequency hopping sequences and channel mapping tables to the second LE-Audio device through a configuration message. The second LE-Audio device updates its local frequency hopping synchronizer and channel selector according to the received parameters.

[0032] Taking the connection between a smartphone and a true wireless stereo left earphone as an example. The smartphone acts as the first LE-Audio device, and the left earphone as the second LE-Audio device. The smartphone initiates a connection isochronous group establishment request, carrying isochronous interval parameters of 7.5 milliseconds, 10 milliseconds, and 15 milliseconds in the message. The left earphone selects an isochronous interval parameter of 10 milliseconds based on its own audio playback latency requirements. The smartphone generates an access address 0x12345601 for the first connection isochronous stream and an access address 0x12345602 for the second connection isochronous stream. The smartphone calculates that the duration of a single sub-event is 5 milliseconds, sets the anchor offset of the second connection isochronous stream to 2500 microseconds, configures the transmission mode of the two connection isochronous streams as interleaved mode, and generates two different frequency hopping sequences to be assigned to the two connection isochronous streams respectively. After receiving all configuration parameters, the left earphone completes the establishment of the connection isochronous group. At this point, two parallel audio data transmission paths are established between the smartphone and the left earphone. Each path contains a 5-millisecond sub-event within a 10-millisecond equal time interval. The two sub-events are arranged alternately on the timeline and are transmitted using different physical channels.

[0033] After the first and second LE-Audio devices complete all parameter configurations, they enter the isochronous stream synchronization phase. The first LE-Audio device sends a synchronization message in the first sub-event of the first isochronous stream connection. This message carries the start timestamp of the isochronous group and the sequence number of the current sub-event. Upon receiving the synchronization message, the second LE-Audio device aligns its local clock with the start timestamp and calculates the accurate time of the sub-event based on its sequence number. The first LE-Audio device then sends the same synchronization message in the first sub-event of the second isochronous stream connection, ensuring that the time synchronization accuracy of the two isochronous streams remains at the microsecond level. After synchronization is complete, the multipath audio transmission channel between the first and second LE-Audio devices is officially established, and audio data transmission can begin.

[0034] In one embodiment, step S2, in which the first LE-Audio device performs encoding processing on the raw audio data to generate audio data packets conforming to the BLE-Audio specification, and splits the audio data packets into N data fragments, where N is an integer greater than or equal to 2, includes: S21, Acquire raw audio data and extract acoustic features; S22, calculates the importance coefficient of audio frames using a lightweight neural network; S23, dynamically allocate coding bits according to importance coefficient to complete encoding; S24 generates standard audio data packets and splits them into multiple data fragments.

[0035] In the specific implementation process, the first LE-Audio device acquires the raw audio data played locally on the terminal. The raw audio data is a standard pulse code modulation digital audio stream with a fixed sampling rate of 48 kHz and a quantization depth of 16 bits. The device performs time-sequential framing of the raw audio data according to the BLEAudio encoding processing standard, with a single frame duration set to 10 milliseconds and a single frame corresponding to 480 raw sampling points. To eliminate the spectral leakage problem caused by frame switching, a 2.5-millisecond data overlap region is maintained between adjacent audio frames. After framing, a Hanning window function is applied to each frame of audio data for smoothing. The length of the window function is perfectly matched with the length of the single frame of audio data to ensure the integrity of the time-domain audio signal.

[0036] A 512-point Fast Fourier Transform is performed on the windowed single-frame audio data to convert the time-domain audio signal into a frequency-domain spectral signal, resulting in 256 sets of effective frequency amplitude data. The device evenly divides the 256 sets of frequency amplitude data into M frequency bands. In this embodiment, M is set to 32. Each frequency band contains 8 sets of continuous frequency amplitude data. The spectral amplitude information within each frequency band is extracted, and a dedicated weighting coefficient is assigned to each frequency band based on the characteristics of human hearing sensitivity. The weighting coefficient is higher for low-frequency hearing-sensitive bands and lower for high-frequency hearing-insensitive bands, achieving accurate extraction of acoustic features.

[0037] The spectral flux of a single frame of audio is calculated using the formula: Ft=Σ(Wk·|X(t,k)|), where k=1 to M, Wk represents the weighting coefficient of the k-th frequency band, X(t,k) represents the spectral amplitude of the k-th frequency band in the t-th frame, and M represents the total number of frequency bands. In this embodiment, the weighting coefficients of the 32 frequency bands are configured hierarchically: Wk is 1.2 for frequency bands 1 to 8, 1.0 for frequency bands 9 to 16, 0.8 for frequency bands 17 to 24, and 0.6 for frequency bands 25 to 32. The measured spectral amplitude of a single frame of audio was selected as the calculation parameter. The spectral amplitudes of the first to the 32nd frequency bands were 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1, 0.1, 0.1, 0.05, 0.05, 0.05, 0.05, 0.02, 0.02, 0.02, 0.02, 0.01, 0.01, 0.01, 0.01, 0.005, 0.005, 0.005, 0.005, 0.002, 0.002, 0.002, 0.002. The calculation is performed by adding up the results of each group according to the formula: the result of the calculation of the first 8 groups of high-frequency sensitive bands is 1.2×0.8+1.2×0.7+1.2×0.6+1.2×0.5+1.2×0.4+1.2×0.3+1.2×0.2+1.2×0.1=4.32; the result of the calculation of the middle 8 groups of intermediate frequency bands is 1.0×0.1+1.0×0.1+1.0×0.05+1.0×0.05+1.0×0.05+1.0×0.05=0.4; the result of the calculation of the last 16 groups of low-frequency insensitive bands is 0.8×0.02×4+0.6×0.01×4+0.6×0.005×4+0.6×0.002×4=0.1348. The results of all frequency band operations are summed to obtain the spectral flux of the current audio frame, Ft=4.8548. The spectral centroid feature of the current audio frame is extracted simultaneously. The spectral centroid is calculated by the weighted average of the frequency band center frequency and the spectral amplitude. It is used to characterize the brightness and darkness of the audio timbre. Together with the spectral flux, it forms the acoustic feature vector of the audio frame, which serves as the input data for the neural network.

[0038] The device extracts spectral flux and spectral centroid to form a two-dimensional acoustic feature vector, which is then input into a pre-defined lightweight neural network. This network strictly adheres to a network structure, including an input layer, convolutional layers, recurrent layers, and an output layer. The convolutional layers are responsible for extracting local correlation features of the acoustic features, while the recurrent layers are responsible for extracting temporal context features of consecutive audio frames, adapting to the temporal characteristics of the audio signal. The neural network performs layer-by-layer operations on the input acoustic feature vector, ultimately outputting the unnormalized raw importance score r of the current audio frame. tTo ensure that the importance coefficients are uniformly constrained within the range of 0 to 1, this step uses a normalization formula: αt = Sigmoid((rt - μr) / σr), where μr is the mean of the original importance scores of multiple consecutive frames, σr is the standard deviation of the original importance scores of multiple consecutive frames, and Sigmoid is a fixed preset activation function, with the function expression being Sigmoid(x) = 1 / (1 + e^(-x)). In this embodiment, the original importance scores of 5 consecutive frames of audio data are selected for batch normalization calculation, and the corresponding rt values ​​for the 5 frames are 2.3, 1.8, 3.1, 0.9, and 2.7, respectively.

[0039] First, calculate the mean score μr. The calculation process is μr = (2.3 + 1.8 + 3.1 + 0.9 + 2.7) ÷ 5 = 10.8 ÷ 5 = 2.16. Then, calculate the standard deviation σr. First, calculate the squared deviation of the score from the mean frame by frame. The results are (2.3 - 2.16)² = 0.0196, (1.8 - 2.16)² = 0.1296, (3.1 - 2.16)² = 0.8836, (0.9 - 2.16)² = 1.5876, and (2.7 - 2.16)² = 0.2916. Sum all the squared deviations to get the total variance of 2.912. Divide this by the number of frames and take the square root to get σr = √(2.912 ÷ 5) = √0.5824 ≈ 0.763.

[0040] The importance coefficient αt is calculated by substituting each frame into the normalization formula. Taking the first frame with rt=2.3 as an example, the standardized deviation (2.3-2.16)÷0.763≈0.183 is first calculated, and then the Sigmoid function is used to calculate 1 / (1+e^(-0.183))≈0.546, resulting in an importance coefficient of 0.546 for this frame. Following the same calculation logic, the importance coefficients for the remaining four frames are 0.461, 0.654, 0.286, and 0.607, respectively. All values ​​satisfy the constraint 0≤αt≤1.

[0041] After calculating the importance coefficients for each audio frame, the device dynamically allocates the number of encoding bits for each audio frame according to the encoding bit allocation formula: Bt=Bbase+αt×(Bmax-Bbase), where Bbase is the preset base number of encoding bits and Bmax is the preset maximum number of encoding bits. In this embodiment, Bbase is fixed at 32 bits and Bmax at 128 bits, with a fixed bit difference of 96 bits. The corresponding importance coefficients are substituted into each frame to complete the calculation. For the first frame, Bt=32+0.546×96≈84 bits; for the second frame, Bt=32+0.461×96≈76 bits; for the third frame, Bt=32+0.654×96≈95 bits; for the fourth frame, Bt=32+0.286×96≈60 bits; and for the fifth frame, Bt=32+0.607×96≈90 bits.

[0042] The device uses a low-complexity communication codec adapted to BLE-Audio to encode each frame according to the number of encoded bits allocated to each frame. Audio frames with higher importance coefficients are allocated more encoded bits to retain richer audio details, while less important audio frames have fewer encoded bits to reduce data redundancy. After encoding, all time-series audio frames are concatenated sequentially, and a standard data packet header defined by the BLE-Audio specification is added. The header contains standardized fields such as audio stream identifier, encoding format, data packet length, and audio packet sequence number, ultimately generating a complete and compliant audio data packet.

[0043] For the generated complete audio data packet, the device performs uniform fragmentation processing. The fragmentation process follows a fixed byte granularity. In this embodiment, the fixed payload length of a single fragment is set to 256 bytes, and the total payload length of the generated audio data packet is 1200 bytes. According to the fixed granularity, it can be evenly divided into 5 groups of data fragments, which satisfies the quantity constraint of N≥2. The payload length of the first four groups of data fragments is 256 bytes, and the remaining payload length of the last group is 176 bytes.

[0044] To ensure the receiving end can accurately identify the ownership and location information of data fragments, the device adds a standardized fragment header to each group of data fragments. The fragment header contains the unique sequence number of the current audio data packet, the fragment index, and the total number of fragments N. The fragment index starts from 0 and increments sequentially. The indices of the 5 groups of fragments correspond to 0, 1, 2, 3, and 4 respectively. All fragment headers uniformly contain the total number of fragments 5, and the fragment header occupies a fixed 4 bytes. After the header encapsulation is completed, the total length of the first four complete groups of fragments is 260 bytes, and the total length of the last group of fragments is 180 bytes. All fragment structures are standardized and ordered, and can be directly used for multi-path redundant transmission.

[0045] In the actual audio transmission scenario between smartphones and true wireless earbuds, the smartphone, as the primary LE-Audio device, continuously collects and processes raw music audio data. Through the complete formula calculation process of the above-mentioned spectral flux, batch normalization, and dynamic bit allocation, it adaptively completes differentiated encoding, generates compliant audio data packets, and evenly splits them into multiple data fragments.

[0046] In one embodiment, the first LE-Audio device calculates the hash value of each data fragment, uses the hash values ​​of the N data fragments as leaf nodes, constructs a hierarchical hash tree by calculating the hash values ​​of the parent nodes layer by layer, obtains the root hash value of the hierarchical hash tree, and generates the transmission metadata of the audio data packet, wherein the transmission metadata includes at least the audio packet sequence number and the number of fragments N. Step S3 includes: S31, Calculate the hash value of each data shard to generate leaf nodes; S32, constructing a hierarchical hash tree by merging layers from bottom to top; S33, calculate the unique root hash value of the hash tree; S34, generate transport metadata containing sequence numbers and the number of fragments.

[0047] In the specific implementation process, the first LE-Audio device reads the N ordered data fragments generated in step S2. In this embodiment, N is 5, corresponding to five sets of data fragments with indices of 0, 1, 2, 3, and 4, respectively denoted as D0, D1, D2, D3, and D4. The device calls a preset lightweight hash function, which is adapted to the embedded computing resources of Bluetooth Low Energy audio devices and can quickly generate fixed-length hash digests from fixed-length binary data fragments for use in the node calculation of a hierarchical hash tree.

[0048] The device generates all the leaf nodes at the bottom layer of the hash tree according to the leaf node calculation formula. The leaf node calculation formula is H(0,i)=Hash(Di), where H(0,i) represents the hash value of the i-th leaf node at the 0th level of the hash tree, Di represents the data fragment with the index i, and Hash represents the preset lightweight hash function. The value range of i corresponds to 0 to N-1.

[0049] For the five data segments in this embodiment, the hash values ​​of the leaf nodes are calculated sequentially by substituting them into the formula, resulting in the hash values ​​H(0,0), H(0,1), H(0,2), H(0,3), and H(0,4) of the five leaf nodes at level 0. Each of the five leaf nodes corresponds one-to-one with one of the five data segments. All leaf node hash values ​​are unique and complete, fully recording the original data characteristics of all segments of the current audio data packet. Any change to the data in any data segment will uniquely change the hash value of the corresponding leaf node. After the leaf nodes are constructed, the device merges nodes layer by layer from bottom to top and calculates the hash value of the upper-level nodes based on the parent node iterative calculation formula. The parent node iterative calculation formula is H(k+1,j)=Hash(H(k,2j)‖H(k,2j+1)), where H(k,2j) and H(k,2j+1) are the hash values ​​of two adjacent child nodes at the k-th level, and H(k+1,j) is the hash value of the parent node at the (k+1)-th level generated by merging the two child nodes. The symbol "‖" represents the concatenation and merging operation of two segments of binary data. The layer-by-layer construction operation of the hierarchical hash tree is completed based on all valid nodes in the current level. Each layer of nodes generates an upper-level parent node according to the pairwise pairing rule. When the number of nodes in a single layer is even, all nodes can be paired to complete the operation. When the number of nodes in a single layer is odd, the independent node at the end of the layer autonomously copies its own hash value to complete the pairing, ensuring that each layer of nodes can fully participate in the upper-level node operation, adapting to any number of shards greater than or equal to 2.

[0050] In this embodiment, level 0 contains five leaf nodes, which is an odd number of nodes. When performing node operations on the first level, the first four nodes are paired and operated on in pairs. The fifth node at the end copies itself to complete the pairing. H(0,0) and H(0,1) are concatenated and input into the hash function to calculate the hash value H(1,0) of the first parent node in the first level. H(0,2) and H(0,3) are concatenated and input into the hash function to calculate the hash value H(1,1) of the second parent node in the first level. H(0,4) is concatenated with its own copied hash value and input into the hash function to calculate the hash value H(1,2) of the third parent node in the first level. After the operation is completed, a total of three valid nodes are generated in the first level of the hash tree.

[0051] Continue iterating upwards to complete the second-level node operation. The three nodes in the first level are also an odd number. Using the same pairing rules, concatenate H(1,0) and H(1,1) to generate the hash value H(2,0) of the first parent node in the second level. Concatenate H(1,2) with its own replicated hash value to generate the hash value H(2,1) of the second parent node in the second level. At this point, the second level of the hash tree has generated two valid nodes.

[0052] The process continues iterating upwards to complete the third-level node calculation. Since the number of nodes in the second level is even, they are directly paired and concatenated. H(2,0) and H(2,1) are concatenated and input into the hash function to finally calculate a unique top-level node hash value. This unique top-level node hash value is the root hash value of the hierarchical hash tree corresponding to the current audio data packet, completing the construction of the entire hash tree. The entire hash tree is interconnected layer by layer from bottom to top. The values ​​of all upper-level nodes are derived from the actual data fragments in the lower layers, and the root hash value aggregates the feature information of all data fragments in the entire packet.

[0053] After the root hash value is calculated, the first LE-Audio device synchronously generates the transmission metadata corresponding to the current audio data packet. The transmission metadata is structured binary data adapted to the BLE-Audio transmission protocol format, and its core includes two core parameters: the audio packet sequence number and the number of fragments N. The audio packet sequence number is a unique identifier that increments with each packet, used to distinguish different audio data packets sent consecutively by the device, avoiding data corruption and matching errors during multi-packet transmission. The number of fragments N is fixedly filled with the actual total number of fragments of the current audio data packet. In this embodiment, it is uniformly filled with the value 5, used to inform the receiving end of the size of the leaf nodes of the current hash tree, providing a benchmark parameter for the receiving end to reconstruct the hash tree locally and verify the fragment integrity. The transmission metadata can be extended to carry auxiliary parameters such as encoding format and frame duration according to transmission adaptation requirements. The core parameters always retain the audio packet sequence number and the number of fragments N, ensuring that the basic verification logic of the receiving end can be executed stably. The device completes the data association binding between the generated root hash value and the structured transmission metadata, so that the unique root hash value corresponds one-to-one with the metadata parameters of the current audio data packet, forming complete data packet verification benchmark information.

[0054] In a specific implementation, in one embodiment, the first LE-Audio device sends the data fragments through the at least two isochronous streams, wherein step S4, in which the same data fragment is transmitted via at least two different isochronous streams, includes: S41, synchronously distribute all data fragments to multiple isochronous streams; S42 uses an interleaved mode to schedule the sub-events of the isochronous streams of each connection; S43 transmits redundant data fragments through different physical channels; S44, retain fragment identification information for matching processing at the receiving end.

[0055] In the specific implementation process, the first LE-Audio device reads all data fragments encapsulated in step S2. In this embodiment, there are five ordered data fragments, with fragment indices of 0, 1, 2, 3, and 4, corresponding to the split data of the complete audio data packet. The device retrieves multiple parallel connection isochronous streams configured in step S1. This embodiment adopts a dual-connection isochronous stream architecture, namely, the first connection isochronous stream and the second connection isochronous stream. The two connection isochronous streams share a unified isochronous interval parameter and have independent air interface transmission resources and sub-event scheduling windows, which can synchronously execute parallel transmission of audio data. The device performs global redundancy allocation processing on all data fragments. The allocation rule follows the single-fragment multi-path redundancy transmission mechanism. The data fragment corresponding to each index is synchronously allocated to all established connection isochronous stream sending queues. There are no data fragments that are only transmitted in a single stream. After the five sets of data fragments are completely copied, they are stored in the sending buffer queue of the first connection isochronous stream and the sending buffer queue of the second connection isochronous stream, respectively. The fragment index order in the two queues is completely consistent with the original audio data packet fragment order, ensuring the consistency of data timing in multi-path transmission.

[0056] Both connections operate in interleaved transmission mode, with the device link layer scheduler using sub-events as the smallest scheduling granularity to control the alternating air interface occupancy of multiple streams. Within the same time interval, the two connections' equal-time streams form interleaved sub-event windows on the time axis based on preset anchor point time offsets. Sub-events of different connections' equal-time streams will not overlap in time, avoiding air interface resource conflicts and signal interference problems caused by concurrent multi-stream transmission.

[0057] Within the first valid sub-event window of the current isochronous connection, the first connection reads five data fragments from its own buffer queue sequentially, encapsulates them into link-layer data packets in ascending order of index, and completes over-the-air transmission. After the first connection isochronous flow completes a single sub-event transmission, the scheduler immediately switches transmission authority to the second connection isochronous flow. Within the corresponding offset sub-event window, the second connection isochronous flow, with the exact same fragmentation order and encapsulation format, sequentially transmits the five redundant data fragments over the air. The two connection isochronous flows correspond to different physical channel resources. The channel frequency hopping sequence and channel mapping table are independent of each other. The same data fragment is synchronously transmitted redundantly on two different physical channels. The independent transmission characteristics of different channels effectively avoid fragment loss or data errors caused by sudden interference, channel fading, or frequency band congestion on a single channel. An anomaly in transmission on one path will not affect the fragment transmission validity of the other path, thus improving the fault tolerance of multi-path transmission at the link level.

[0058] During dual-path transmission, all data fragments retain their header identification information. Each redundant fragment copy carries the same audio packet sequence number, fragment index, and total fragment count. The second LE-Audio receiving device can use this unified identification information to match identical fragment copies from different isochronous streams. The multi-path transmission process relies entirely on a unified isochronous interval timing benchmark. The transmission cycles of the two isochronous streams are completely synchronized, and the time loss from alternating sub-event scheduling is controlled within the microsecond-level synchronization accuracy of the Bluetooth protocol, meeting the core requirements of LE-Audio's low-latency audio transmission. After all data fragments have completed dual-path redundant transmission, the device retains the fragment index record and transmission timing record of the current transmission queue, and uses the check record identifier to complete the data transmission closed-loop verification.

[0059] In real-world scenarios where smartphones transmit audio data to true wireless stereo headphones, the smartphone, acting as the primary LE-Audio device, synchronously distributes all five segments of the split audio data to two parallel, isochronous streams. Leveraging an interleaved polling sub-event scheduling mechanism and independent physical channel resources, dual-path redundant transmission of each data segment is achieved. Even if any link experiences brief signal interference, channel switching, or air interface preemption anomalies, the redundant segments on the other link can still be delivered to the receiving end normally, effectively solving the problems of easy stuttering, packet loss, and poor reliability in traditional single-path audio transmission.

[0060] In one embodiment, step S5, where the first LE-Audio device associates the root hash value with the transmission metadata and writes it into a local anti-tampering log to generate a corresponding verification record identifier, includes: S51, associate and bind the root hash value with the transmitted metadata; S52, generates verification record identifiers based on chained hash algorithm; S53, append the verification record to the local anti-tampering log; S54, synchronous output check record identifier is used for transmission verification.

[0061] In the specific implementation process, the first LE-Audio device retrieves the root hash value of the hierarchical hash tree and the transmission metadata of the current audio data packet generated in step S3. The transmission metadata includes the audio packet sequence number and the number of fragments N of the current audio packet. In this embodiment, the number of fragments N is set to 5. The device structurally associates and binds the root hash value with the transmission metadata, so that a unique set of root hash values ​​corresponds to unique audio data packet metadata information, accurately locking the fragment characteristics and packet attribute information of the current whole audio data packet, forming the basic verification dataset of the current audio data packet.

[0062] The local anti-tampering log is a linear, append-only data structure maintained exclusively in the device's local storage area. This storage area features hardware read / write permission locking. Records already written to the log cannot be deleted, overwritten, or tampered with; only appending new records to the end is supported. The log stores verification records sequentially according to the transmission time of the audio data packets. Each record corresponds to an independent audio data packet, forming a sequentially continuous chain of verification records, providing traceable integrity audit evidence for audio transmission data.

[0063] The device constructs a complete verification record for the current audio data packet and calculates the corresponding verification record identifier based on the chained verification identifier generation formula. The calculation formula is: Rid(i) = Hash(Root(i)‖Meta(i)‖Rid(i-1)), where Rid(i) represents the verification record identifier corresponding to the i-th audio data packet, Root(i) represents the root hash value of the hierarchical hash tree of the current audio data packet, Meta(i) represents the transmission metadata of the current audio data packet, Rid(i-1) represents the verification record identifier corresponding to the previous audio data packet, ‖ represents the concatenation and merging operation of multiple binary data segments, and Hash is a preset lightweight hash function consistent with the hash tree construction process to ensure the uniformity of the device's global hash operation rules. The first audio data packet has no prior historical verification record. The device defaults to using an initial null value as the pre-concatenation parameter for the first data packet to complete the calculation of the first verification record identifier. Starting from the second audio data packet, the verification record identifier of each data packet is associated with the identifier information of the previous data packet, so that the verification records of the audio data packets form a chain binding relationship. If any record is tampered with, it will cause all verification record identifiers to change as a whole, thus eliminating the possibility of tampering with a single record from the algorithm level.

[0064] In the audio transmission scenario of this embodiment, the currently processed audio data packet is a valid data packet within the sequence. The device extracts the root hash value Root(i) of the current packet, the transmission metadata Meta(i) containing the packet sequence number and the number of fragments (5), and retrieves the historical verification record identifier Rid(i-1) generated from the previous frame of audio data packets. The three data segments are then concatenated in a fixed order to obtain the complete original data of the current verification record. The concatenated complete data is input into a preset hash function for irreversible hashing to generate a fixed-length, unique, and irreversible current verification record identifier Rid(i). After the verification record identifier calculation is completed, the device binds the current verification record and the current verification record identifier one-to-one and stores them in the local anti-tampering log by appending them. A single complete verification record entry contains the root hash value of the current audio data packet, the transmission metadata, the source information of the previous verification record identifier, and the current verification record identifier. All entries are arranged linearly according to the data packet sending sequence, without disorder, overwriting, or deletion, and continuously iterate to form a closed-loop chain verification log. The first LE-Audio device synchronously sends the generated current verification record identifier to the second LE-Audio device along with the data fragment. The verification record identifier can be carried in the extended header field of the data fragment, or it can be transmitted through a separate control sub-event. After receiving the verification record identifier, the second LE-Audio device can match the corresponding verification record in the local anti-tampering log of the first LE-Audio device according to the unique identifier index, extract the base root hash value and metadata information, and use them to compare with the locally reconstructed hash tree data to complete the data fragment integrity verification after multi-path transmission.

[0065] In actual transmission scenarios between smartphones and true wireless stereo headphones, the smartphone continuously performs chain-like verification calculations on each encoded and segmented audio data packet, generating a unique verification record identifier for each packet and storing it in a local anti-tampering log. During multi-path redundant transmission, even if some path fragments experience abnormalities such as data loss, bit errors, or interference distortion, the headphones can still trace the baseline verification data through the verification record identifier, accurately identify abnormal fragments, ensure the accuracy and traceability of audio data verification under the multi-path transmission architecture, and further improve the stability and reliability of LE-Audio audio transmission.

[0066] Reference Figure 2 Here is a structural block diagram of a multipath audio data transmission device between LE-Audio devices according to an embodiment of the present invention, comprising: A connection establishment unit is used to establish an isochronous group with a second LE-Audio device based on the Bluetooth Low Energy Audio protocol; The encoding fragmentation unit is used to perform encoding processing on the original audio data, generate audio data packets conforming to the BLE-Audio specification, and split the audio data packets into N data fragments, where N is an integer greater than or equal to 2; The hash tree construction unit is used to calculate the hash value of each data fragment, with the hash values ​​of the N data fragments as leaf nodes, and to construct a hierarchical hash tree by calculating the hash values ​​of the parent nodes layer by layer, thereby obtaining the root hash value of the hierarchical hash tree and generating the transmission metadata of the audio data packet. A multipath transmission unit is used to transmit the data fragments respectively through the at least two connection isochronous streams, wherein the same data fragment is transmitted via at least two different connection isochronous streams; The log writing unit is used to associate the root hash value with the transmission metadata and write it into the local anti-tampering log to generate a corresponding verification record identifier.

[0067] For the specific implementation of each module in the above device example, please refer to the above method embodiments, which will not be repeated here.

[0068] In summary, this invention achieves low-latency, high-reliability, and traceable Bluetooth Low Energy (BLE) audio data transmission by constructing a parallel transmission architecture based on multiple connection isochronous streams between LE-Audio devices. This architecture integrates fragment-level redundant distribution, dynamic adaptive coding, hierarchical hash tree integrity verification, and chain-based anti-tampering log technology. Within the BLE protocol framework, by configuring connection isochronous groups containing multiple independent connection isochronous streams and employing interleaved transmission modes and differentiated physical channel allocation mechanisms, multiple transmission paths alternately occupy air interface resources at the sub-event granularity, effectively avoiding channel interference and resource conflicts inherent in single-path transmission. This significantly improves link fault tolerance while maintaining low transmission latency. Regarding the quality requirements of audio data transmission, by extracting the spectral flux and spectral centroid features of audio frames, a lightweight neural network is used to calculate the importance coefficient of audio frames and dynamically allocate coding bits, achieving this without increasing the total transmission bandwidth. The audio encoding quality was optimized, and the encoded audio data packets were split into multiple independent data fragments, laying the foundation for multi-path redundant transmission. To ensure the data integrity of multi-path transmission, a hierarchical hash tree technology was adopted. The hash tree was constructed from the bottom up using the hash value of each data fragment as the leaf node. The feature information of the entire data packet was aggregated through a unique root hash value. Combined with a chain-like anti-tampering log mechanism, the root hash value was associated and bound with the transmission metadata to generate an immutable verification record identifier, realizing integrity verification and traceability auditing of the entire audio data transmission process. During the data transmission phase, each data fragment was synchronously distributed to multiple connections and isochronous streams for redundant transmission. The same data fragment was sent independently through different physical channels. When a sudden interference occurs on one path and a fragment is lost, the redundant fragments on other paths can still be delivered to the receiving end normally, solving the problems of easy packet loss and stuttering in traditional single-path LE-Audio transmission.

[0069] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0070] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0071] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0072] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for multipath transmission of audio data between LE-Audio devices, characterized in that, Includes the following steps: The first LE-Audio device and the second LE-Audio device establish a connection isochronous group based on the Bluetooth Low Energy Audio protocol. The connection isochronous group is configured to include at least two independent connection isochronous streams, forming multiple parallel audio data transmission paths, and each connection isochronous stream has at least one sub-event in each isochronous interval. The first LE-Audio device performs encoding processing on the raw audio data to generate audio data packets conforming to the BLE-Audio specification, and splits the audio data packets into N data fragments, where N is an integer greater than or equal to 2; The first LE-Audio device calculates the hash value of each data segment, uses the hash values ​​of the N data segments as leaf nodes, constructs a hierarchical hash tree by calculating the hash values ​​of the parent nodes layer by layer, obtains the root hash value of the hierarchical hash tree, and generates the transmission metadata of the audio data packet. The transmission metadata includes at least the audio packet sequence number and the number of segments N. The first LE-Audio device sends the data fragments through the at least two isochronous streams, wherein the same data fragment is transmitted via at least two different isochronous streams. The first LE-Audio device associates the root hash value with the transmission metadata and writes it into the local anti-tampering log to generate a corresponding verification record identifier.

2. The method for multipath transmission of audio data between LE-Audio devices according to claim 1, characterized in that, The first LE-Audio device performs encoding processing on the raw audio data to generate audio data packets conforming to the BLE-Audio specification, including: The first LE-Audio device acquires raw audio data and extracts acoustic features from the raw audio data, the acoustic features including spectral flux F. t and the spectral centroid C t ; Where the spectral flux F t Satisfy: F t =Σ(W k ·|X(t,k)|), k=1 to M, W k X(t,k) is the weighting coefficient of the k-th frequency band, X(t,k) is the spectral amplitude of the k-th frequency band in the t-th frame, and M is the total number of frequency bands; The acoustic features are input into a preset lightweight neural network, which then outputs the importance coefficient α of each audio frame in the original audio data. t The importance coefficient α t Satisfying 0≤α t ≤1, the lightweight neural network is configured as an end-to-end structure with the acoustic features as input and the importance coefficient as output; Based on the importance coefficient α t Assign coding bits to each audio frame, and assign B coding bits to the t-th frame. t =B base +α t ×(B max -B base ), where B base Based on the number of basic encoding bits, B max To maximize the number of encoded bits, the original audio data is encoded to generate the audio data packet conforming to the BLE-Audio specification.

3. The method for multipath transmission of audio data between LE-Audio devices according to claim 2, characterized in that, The step of inputting the acoustic features into a preset lightweight neural network and having the lightweight neural network output the importance coefficient αt of each audio frame in the original audio data includes: The lightweight neural network includes an input layer, at least one convolutional layer, at least one recurrent layer, and an output layer, wherein the convolutional layer is configured to extract local features from the acoustic features, and the recurrent layer is configured to extract temporal context information from the feature sequence output by the convolutional layer. The acoustic features of frame t are input into the lightweight neural network, and the lightweight neural network outputs the original importance score r corresponding to that frame. t Original importance score r t For unnormalized real numbers; According to the r t The importance coefficient α is calculated using the BatchNorm normalization function, along with the total number of audio frames T. t , satisfying: α t =Sigmoid((r t -μ r ) / σ r ) Where μ r For the r t The mean, σ r For the r t The standard deviation of α, where sigmoid is the preset activation function, such that α t Satisfying 0≤α t ≤1.

4. The method for multipath transmission of audio data between LE-Audio devices according to claim 1, characterized in that, The step of constructing a hierarchical hash tree by calculating the hash values ​​of parent nodes layer by layer, and obtaining the root hash value of the hierarchical hash tree, includes: The N data fragments are used as the N leaf nodes of the 0th layer, and the hash value H(0,i) of each leaf node in the 0th layer is Hash(D). i ), where D i This represents the i-th data fragment, where 0 ≤ i ≤ N-1, and Hash represents the preset hash function; For a node at level k, the hash value of the parent node at level k+1 is calculated using the hash values ​​of two adjacent nodes as input, satisfying H(k+1,j)=Hash(H(k,2j)‖H(k,2j+1)), where k≥0 and j≥0; When only a unique parent node hash value is calculated for the (k+1)th layer, the unique parent node hash value is used as the root hash value of the hierarchical hash tree.

5. The method for multipath transmission of audio data between LE-Audio devices according to claim 1, characterized in that, The step of the first LE-Audio device sending the data fragments through the at least two isochronous streams includes: The first LE-Audio device allocates the N data fragments to the at least two connection isochronous streams, and the same data fragment is allocated to at least two different connection isochronous streams; At least one sub-event sends the allocated data fragment within the isochronous interval of each connection isochronous stream, wherein the isochronous intervals of different connection isochronous streams are staggered on the time axis; The transmission mode of the connected isochronous streams is configured as an interleaved mode, in which the sub-events of the at least two connected isochronous streams are sent alternately in a polling manner.

6. The method for multipath transmission of audio data between LE-Audio devices according to claim 1, characterized in that, The step of the first LE-Audio device associating the root hash value with the transmission metadata and writing it into the local anti-tampering log to generate a corresponding verification record identifier includes: The first LE-Audio device combines the root hash value corresponding to the current audio data packet, the transmission metadata, and the previous verification record identifier into the current verification record; A preset hash function is performed on the current verification record to obtain the current verification record identifier, which satisfies: Rid(i) = Hash(Root(i) || Meta(i) || Rid(i-1)); Where Rid(i) is the verification record identifier corresponding to the current audio data packet, Root(i) is the root hash value corresponding to the current audio data packet, Meta(i) is the transmission metadata corresponding to the current audio data packet, Rid(i-1) is the previous verification record identifier, || represents the data connection, and Hash is the preset hash function; The current verification record and its identifier are written into the local anti-tampering log. The local anti-tampering log is a linear data structure that only appends to writes. The second LE-Audio device verifies the integrity of the received data fragments by obtaining the verification record identifier and the corresponding verification record in the local anti-tampering log.

7. The method for multipath transmission of audio data between LE-Audio devices according to claim 1, characterized in that, The step of configuring the connected isochronous group to include at least two independent connected isochronous streams to form multiple parallel audio data transmission paths includes: The first LE-Audio device and the second LE-Audio device establish the connection isochronous group based on the Bluetooth Low Energy Audio protocol. The connection isochronous group includes a first connection isochronous stream and a second connection isochronous stream. The first connection isochronous stream includes a first access address and a first channel resource, and the second connection isochronous stream includes a second access address and a second channel resource. The first connection isochronous stream and the second connection isochronous stream share the same isochronous interval parameter. Within the same time interval, there is a preset offset between the anchor time of the first connected isochronous stream and the anchor time of the second connected isochronous stream, and the first channel resource and the second channel resource correspond to different physical channels respectively; The sub-events of the first connection isochronous stream and the sub-events of the second connection isochronous stream are arranged alternately on the time axis. The transmission modes of the first connection isochronous stream and the second connection isochronous stream are both configured as interleaved, forming the multiple audio data transmission paths that are logically parallel.

8. A multipath transmission device for audio data between LE-Audio devices, characterized in that, The method for multipath transmission of audio data between LE-Audio devices according to any one of claims 1-7, wherein the multipath transmission apparatus for audio data between LE-Audio devices comprises: A connection establishment unit is used to establish a connection isochronous group with a second LE-Audio device based on the Bluetooth Low Energy Audio protocol; The encoding fragmentation unit is used to perform encoding processing on the original audio data, generate audio data packets conforming to the BLE-Audio specification, and split the audio data packets into N data fragments, where N is an integer greater than or equal to 2; The hash tree construction unit is used to calculate the hash value of each data fragment, with the hash values ​​of the N data fragments as leaf nodes, and to construct a hierarchical hash tree by calculating the hash values ​​of the parent nodes layer by layer, thereby obtaining the root hash value of the hierarchical hash tree and generating the transmission metadata of the audio data packet. A multipath transmission unit is used to transmit the data fragments respectively through the at least two connection isochronous streams, wherein the same data fragment is transmitted via at least two different connection isochronous streams; The log writing unit is used to associate the root hash value with the transmission metadata and write it into the local anti-tampering log to generate a corresponding verification record identifier.

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 steps of the LE-Audio device-to-device audio data multipath transmission method as described in 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 the processor, it implements the steps of the LE-Audio device-to-device audio data multipath transmission method as described in any one of claims 1 to 7.