AI voice expansion method and system based on BLE network
By constructing a half-duplex, multi-hop voice pass-through network based on BLE, and combining OPUS encoding and a custom BLE data packet structure, the problems of high cost and limited interaction range in multi-point deployment of traditional AI speakers are solved, achieving low-latency, high-bandwidth voice interaction effects with good naturalness and usability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG LIERDA KESI INTELLIGENT TECH CO LTD
- Filing Date
- 2025-12-30
- Publication Date
- 2026-04-10
AI Technical Summary
Traditional AI smart speakers suffer from limited voice interaction range and high deployment costs, especially when multiple panels are deployed simultaneously, which weakens offline voice recognition capabilities and increases deployment costs.
An AI voice extension method based on BLE network is adopted. A half-duplex, multi-hop voice pass-through network is constructed through AI interaction nodes, user interaction nodes and relay nodes. Voice data is encapsulated using OPUS encoding format and transmitted on BLE link. Combined with 2M PHY physical layer and custom BLE data packet structure, ACK and retransmission mechanism are disabled to achieve low latency and high bandwidth voice transmission.
It achieves low-cost, high-efficiency multi-point voice coverage, with good voice interaction effect and end-to-end latency of less than 300 milliseconds, meeting the real-time requirements of natural dialogue, and defends against replay attacks and man-in-the-middle tampering through a lightweight security mechanism.
Smart Images
Figure CN121838752A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent voice interaction technology, specifically to an AI voice extension method and system based on BLE networks. Background Technology
[0002] Smart speakers are an upgrade from traditional speakers, serving as a tool for home consumers to access the internet via voice commands, such as playing songs, shopping online, or checking the weather forecast. They can also control smart home devices, such as opening curtains, setting refrigerator temperatures, and preheating water heaters. Traditional AI smart speakers feature voice recognition, picking up and responding to audio within a certain range. When connected to the internet, they have semantic recognition, while offline they only support voice recognition. Secondly, distributed voice panels, based on traditional solutions, reduce costs by weakening offline voice recognition capabilities, and deploying multiple panels simultaneously increases the range of voice interaction.
[0003] In summary, while the interaction range of intelligent voice interaction terminals is limited, maintaining a certain level of voice control capability offline can lead to increased deployment costs. Summary of the Invention
[0004] This invention provides an AI voice extension method and system based on BLE network, which has the advantages of good voice interaction effect and low cost.
[0005] This invention provides the following technical solution: an AI voice extension method based on BLE network, comprising at least one AI interaction node, multiple user interaction nodes, and several relay nodes, including the following steps:
[0006] The AI interaction node is used to receive and process voice data packets from the user interaction node, and return the corresponding AI voice response to the user interaction node;
[0007] The user interaction node is used to collect the user's voice and send it to the AI interaction node, while playing the voice response returned by the AI interaction node;
[0008] Specifically, the voice response is encapsulated into a custom BLE data packet in OPUS encoding format and broadcast to the AI interaction node, and the AI voice response returned by the AI interaction node is decoded and played upon receipt;
[0009] The relay node is used to forward voice data packets between the user interaction node and the AI interaction node.
[0010] As an optional solution to the AI voice extension method based on BLE networks described in this invention, the custom BLE data packet includes redefining the BLE extended broadcast PDU structure, comprising:
[0011] Using the BLE 2M PHY physical layer transmission mode, the payload portion of the ADV_EXT_IND broadcast packet is multiplexed into a composite field of the header and the first segment of OPUS voice data;
[0012] Subsequent voice data is transmitted cyclically at fixed time intervals on a preset channel sequence via the AUX_ADV_IND auxiliary packet to avoid 2M PHY adjacent channel interference.
[0013] As an optional solution of the AI voice extension method based on BLE network described in this invention, the user interaction node and the relay node obtain the byte root key from the AI interaction node through a secure pairing process during the initial network configuration phase, and derive a session key based on the root key for MIC verification and optional encryption of the custom BLE data packet in subsequent communication.
[0014] After receiving a voice data packet, the AI interaction node verifies whether its sequence number is monotonically increasing and whether the MIC verification is successful. If either verification fails, the data packet is discarded to prevent replay attacks and man-in-the-middle tampering.
[0015] As an optional solution to the AI speech extension method based on BLE network described in this invention, the payload of ADV_EXT_IND is multiplexed into a composite field of the guide header and the first OPUS speech frame.
[0016] The subsequent voice frames are continuously transmitted on a preset non-adjacent channel sequence via the AUX_ADV_IND auxiliary packet to form voice data;
[0017] The receiving node receives complete voice data through synchronous scanning and channel frequency hopping prediction mechanisms, without relying on the message segmentation and reassembly logic of BLE Mesh;
[0018] The retransmission and acknowledgment mechanism of the BLE connection layer is disabled during the voice data transmission process to maintain low latency.
[0019] As an optional solution to the AI voice extension method based on BLE network described in this invention, the relay node performs transparent forwarding only based on the target address field in the custom BLE data packet.
[0020] As an optional solution to the AI voice extension method based on BLE network described in this invention, the custom BLE data packet includes a source address field SRC_ADDR, a destination address field DST_ADDR, a 16-bit sequence number SEQ_H / L, and a message integrity check code calculated based on the session key.
[0021] The present invention also provides a system for applying an AI speech extension method based on a BLE network, comprising:
[0022] AI interaction node, which is used to receive user voice data, generate AI voice response, and return the AI voice response to the corresponding user interaction node along the original path;
[0023] The user interaction node is used to collect far-field voice from users, encapsulate the voice into a custom BLE data packet in OPUS encoding format and broadcast it to the AI interaction node, and decode and play the AI voice response returned by the AI interaction node.
[0024] Each relay node contains only a BLE communication module and is configured to transparently forward voice data packets according to the target address field in the custom BLE data packet, without performing voice decoding, AI processing, or audio playback.
[0025] The AI interaction node, user interaction node, and relay node together constitute a half-duplex, multi-hop BLE voice pass-through network, and there is only one AI interaction node in the entire network.
[0026] The present invention has the following beneficial effects:
[0027] 1. The AI voice extension method and system based on BLE network can achieve an economical coverage solution with one AI master controller and multiple voice entry points by integrating only low-cost BLE SoC, microphone and speaker into user interaction nodes and relay nodes, without the need for NPU, DSP or high-performance processor, thereby achieving good voice interaction effect and low cost.
[0028] 2. This AI voice extension method and system based on BLE network redefines the BLE extended broadcast PDU structure, combines the 2M PHY physical layer with OPUS audio coding, and achieves near-telephone quality voice transmission on the BLE link. It also adopts end-to-end delay control to meet the real-time requirements of natural dialogue. By disabling ACK and retransmission mechanisms, it avoids voice stuttering caused by packet loss and retransmission, thus improving fluency.
[0029] 3. The AI voice extension method and system based on BLE network fundamentally solves the problem of high-traffic voice transmission by constructing a dedicated voice pass-through channel and adopting directional routing, transparent relay and half-duplex time-division multiplexing. It also clearly distinguishes between Mesh topology and Mesh protocol, which retains the flexibility of multi-hop networking and avoids protocol overhead. Attached Figure Description
[0030] Figure 1 This is a schematic diagram of the packet sending process of the interactive node in this invention.
[0031] Figure 2This is a schematic diagram of the relay node operation process of the present invention.
[0032] Figure 3 This is a schematic diagram of the AI interaction node process of the present invention. Detailed Implementation
[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0034] Example 1
[0035] Please see Figures 1-3 Furthermore, a method for extending AI voice based on BLE network is disclosed, comprising at least one AI interaction node, multiple user interaction nodes, and several relay nodes.
[0036] The AI interaction node is used to receive and process voice data packets from the user interaction node, and return the corresponding AI voice response to the user interaction node.
[0037] The user interaction node is used to collect the user's voice and send it to the AI interaction node, while playing the voice response returned by the AI interaction node.
[0038] Specifically, the voice response is encapsulated in OPUS encoding format into a custom BLE data packet and broadcast to the AI interaction node, and then decoded and played after receiving the AI voice response returned by the AI interaction node.
[0039] Relay nodes are used to forward voice data packets between user interaction nodes and AI interaction nodes.
[0040] AI interaction nodes (A nodes) are typically integrated into the main smart speaker or gateway. They have built-in local or cloud-based AI voice engines, such as NLP+TTS, responsible for speech recognition, semantic understanding, and speech synthesis. They do not actively pick up voices; they only process the voice received from U nodes and return the response along the original path.
[0041] User interaction nodes (U nodes), deployed in various rooms (such as walls, lights, and sockets), are scalable in number and include microphones, speakers, and BLE communication modules (such as ESP32-C2). They are responsible for far-field voice pickup, OPUS encoding and broadcast transmission, as well as receiving, decoding, and playing AI responses. They have no AI processing capabilities and only function as voice pass-through terminals.
[0042] Relay node (R node), optional deployment, used for signal extension. Contains only a BLE communication module, without microphone / speaker connection; transparently forwards voice data packets: directly transmits them according to the destination address (DST_ADDR), without decoding, verification, retransmission, or acknowledgment.
[0043] The entire network forms a half-duplex, multi-hop, single-master voice pass-through architecture.
[0044] Specifically, when a user speaks in a room, the nearest U node picks up the far-field speech through its microphone. After front-end noise reduction, it is OPUS encoded with fixed parameters (16kHz sampling rate, 24kbps bitrate). The U node then encapsulates the OPUS speech frames into a custom BLE extended broadcast packet, which includes:
[0045] SRC_ADDR (source address, i.e., U node ID);
[0046] DST_ADDR (target address, fixed as the address of node A);
[0047] 16-bit serial number (SEQ_H / L, monotonically increasing);
[0048] MIC (Message Integrity Check Code calculated based on session key);
[0049] OPUS voice payload;
[0050] Improve throughput using the BLE 2M PHY physical layer:
[0051] The ADV_EXT_IND PDU is redefined, and its payload is multiplexed into a header and the first OPUS data segment. Subsequent voice frames are transmitted cyclically at fixed intervals on preset non-adjacent channels (such as 0, 2, 4…36) via AUX_ADV_IND auxiliary packets to form a continuous voice chain.
[0052] Disable the ACK and retransmission mechanism of the BLE connection layer to avoid latency accumulation due to packet loss.
[0053] If node U and node A are not within direct connection range, after node R detects a data packet with a destination address of A, it immediately transmits it without performing any protocol processing.
[0054] Node A receives the complete voice stream through synchronous scanning and channel frequency hopping prediction, and verifies it by checking if the sequence number is monotonically increasing (to prevent replay), if the microphone is valid (to prevent tampering), and discarding the stream if verification fails. If verification is successful, the OPUS audio is decoded and sent to the AI engine to generate a response voice.
[0055] The AI response encoding and reverse routing: Node A encodes the generated TTS voice using OPUS, encapsulates it into a custom BLE packet, and sets DST_ADDR to the original U node's SRC_ADDR, enabling a return along the original path. Downlink broadcast and relay response packets are also sent via 2M PHY and extended broadcast mechanism, transparently transmitted through node R, and finally reach the target U node. Local playback: After receiving and verifying the response packet, node U decodes the OPUS audio and plays it through its built-in speaker, completing a full interaction. The entire process is in half-duplex mode, with uplink (U to A) and downlink (A to U) time-division multiplexing to avoid collisions.
[0056] It should be noted that during the initial network configuration phase, the U / R node obtains an 8-byte root key from the A node through secure pairing (such as OOB or PIN code);
[0057] Session key derivation: A session key is generated based on the root key and the node address, and is used for MIC calculation;
[0058] Runtime security is ensured by verifying the MIC and serial number of each packet to guarantee the integrity and freshness of communication.
[0059] The lightweight design avoids the complex NetKey / AppKey system of BLE Mesh, reducing resource consumption.
[0060] In summary, user interaction nodes and relay nodes only need to integrate low-cost BLE chips, microphones, and speakers, without the need for high-performance processors or AI acceleration units. The hardware cost per point can be controlled within RMB 10, which is far lower than that of traditional full-featured AI speakers, achieving cost-effective and efficient multi-point voice coverage.
[0061] By employing the BLE 2M PHY physical layer, OPUS audio encoding (16kHz sampling rate, approximately 24kbps bitrate), and a custom extended broadcast protocol, stable transmission of near-telephone quality voice streams is achieved over the BLE link, with end-to-end latency below 300 milliseconds, meeting the real-time requirements of natural dialogue.
[0062] This solution completely abandons the flooding, message segmentation and reassembly, and ACK retransmission mechanisms of BLE Mesh, instead constructing a dedicated voice pass-through channel. It employs directional routing, transparent relaying, and half-duplex time-division multiplexing strategies to fundamentally solve the challenges of high-bandwidth, low-latency voice transmission. It supports flexible multi-hop networking to expand coverage. Multi-hop voice pass-through is achieved through relay nodes, penetrating walls and covering complex apartment layouts. The relay nodes are low-power, small in size, and easily embedded in various IoT devices, enabling seamless deployment and self-organizing networks.
[0063] A lightweight security mechanism is adopted, which issues an 8-byte root key and derives a session key during the network configuration phase. Combined with a 16-bit monotonically increasing sequence number and message integrity verification (MIC), it effectively defends against replay attacks and man-in-the-middle tampering, while avoiding the computational overhead of the BLEMesh two-layer key system.
[0064] Built on a mature BLE 5.x chip platform (such as ESP32-C2 / C3), it requires no custom RF or protocol stack. User interaction nodes can be integrated into existing terminals such as smart switches, lights, and sockets. AI interaction nodes can be implemented through firmware upgrades to existing smart speakers, reusing existing supply chains and product forms. Users can speak from any location, with the nearest user interaction node picking up the voice and transmitting it to the AI master control. The response voice is then played locally from that node, achieving proximity-based interaction and local echo, avoiding the problem of traditional solutions requiring proximity to the main speaker for activation, and significantly improving the naturalness and usability of voice interaction.
[0065] It is important to note that standard BLE extended broadcast uses AuxPtr to chain auxiliary packets, causing the receiver to hop multiple times and introduce uncontrollable latency (>100ms), failing to meet the <200ms experience requirement for voice interaction. This invention, through scheduled timing and channel sequencing, splits the entire voice frame into continuous broadcast packets and sends them in a fixed channel sequence. The receiver only needs to scan periodically to receive the complete frame without hopping, reducing the measured end-to-end latency to 60ms. Ordinary BLE broadcast cannot meet bandwidth requirements. Redesigning based on extended broadcast avoids channel congestion and reduces development effort. Mesh networks are suitable for low-traffic, multi-node, high-security scenarios and cannot support real-time voice streams. The voice network in this embodiment, based on a mesh network architecture, is specifically redesigned for high traffic and low latency to meet performance requirements. Extended broadcast packets defined in SIG BLE have a data limit; once the limit is reached, the next packet needs to be sent, which impacts voice real-time performance (each packet split causes a delay of over 200ms). Therefore, the extended packet protocol was redefined based on the original protocol.
[0066] It should be noted that, to rigorously verify the technical and commercial advantages of this solution, we conducted 1000 full-scenario stress tests (covering a 100㎡ three-bedroom, one-living room environment, including three room partitions) on key performance indicators. The experimental data is shown in the table below. Test conditions:
[0067] The total cost of this plan is 90 yuan, consisting of 1 A node (15 yuan), 3 U nodes (15 yuan each), and 3 R nodes (10 yuan each).
[0068] The existing solution involves 4 AI speakers (200 yuan each) = a total cost of 800 yuan;
[0069]
[0070] Example 2
[0071] The present invention also provides a system for applying an AI speech extension method based on a BLE network, comprising:
[0072] AI interaction node: The AI interaction node is used to receive user voice data, generate AI voice response, and return the AI voice response to the corresponding user interaction node along the original path.
[0073] The user interaction node is used to collect far-field voice from users, encapsulate the voice into a custom BLE data packet in OPUS encoding format and broadcast it to the AI interaction node, and decode and play the AI voice response returned by the AI interaction node.
[0074] Each relay node contains only a BLE communication module and is configured to transparently forward voice data packets based on the destination address field in the custom BLE data packet, without performing voice decoding, AI processing, or audio playback.
[0075] The AI interaction node, user interaction node, and relay node together form a half-duplex, multi-hop BLE voice pass-through network, and there is only one AI interaction node in the entire network.
[0076] It should be noted that network protocols include:
[0077] Equipment network distribution:
[0078] Connect to the device by scanning the QR code with the mobile app. Send the network root key (8 bytes) and device address (2 bytes) to the device via GATT Service: 0xfb, 0x34, 0x9b, 0x5f, 0x80, 0x00, 0x00, 0x80, 0x00, 0x10, 0x00, 0x00, 0xb0, 0xbb, 0x31, 0x45. After successful setup, the device will return the network configuration status via NOTIFY.
[0079] Message packet format (logic layer):
[0080] ;
[0081] ;
[0082] This example uses the ESP32-C2 Bluetooth module and the ESP-IDF protocol stack. The PDU format of the link layer in Extended Broadcast has been modified. The ADV_EXT_IND field has been redefined according to the protocol of this invention and the AuxPtr field has been omitted. By using a unified agreed transmission and scanning cycle, a rate of about 80 kbps was achieved in a simulated 50-node environment. This is sufficient to transmit audio signals in the network while preserving the voiceprint information.
[0083] This specification also provides a computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform the multiple steps described in the above embodiments. If the constituent modules of the above-described electronic device are implemented as software functional units and sold or used as independent products, they can be stored in the computer-readable storage medium.
[0084] This specification also provides a computer program product, including a computer program that, when executed by a processor, implements the multiple steps described in the above embodiments.
[0085] Where there is no conflict, the technical features in this embodiment and implementation scheme can be combined arbitrarily.
[0086] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes multiple computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center integrating multiple available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital versatile discs (DVDs)), or semiconductor media (e.g., solid-state drives (SSDs)).
[0087] When implemented through hardware or firmware, the aforementioned method flow is programmed into the hardware circuit to obtain the corresponding hardware circuit structure and achieve the corresponding function. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit, whose logic function is determined by the user programming the device. Designers can program a digital system onto a PLD themselves, eliminating the need for chip manufacturers to design and fabricate dedicated integrated circuit chips. Furthermore, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly implemented using "logic compiler" software, similar to the software compiler used in program development. The original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There is not just one HDL, but many. Those skilled in the art should understand that by simply performing some logic programming on the method flow using one of the aforementioned hardware description languages and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.
[0088] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0089] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. An AI speech extension method based on BLE networks, characterized in that, The system comprises at least one AI interaction node, multiple user interaction nodes, and several relay nodes, and includes the following steps: The AI interaction node is used to receive and process voice data packets from the user interaction node, and return the corresponding AI voice response to the user interaction node; The user interaction node is used to collect the user's voice and send it to the AI interaction node, while playing the voice response returned by the AI interaction node; Specifically, the voice response is encapsulated into a custom BLE data packet in OPUS encoding format and broadcast to the AI interaction node, and the AI voice response returned by the AI interaction node is decoded and played upon receipt; The relay node is used to forward voice data packets between the user interaction node and the AI interaction node.
2. The AI speech extension method based on BLE network according to claim 1, characterized in that, The custom BLE data packet redefines the BLE Extended Broadcast PDU structure, including: Using the BLE 2M PHY physical layer transmission mode, the payload portion of the ADV_EXT_IND broadcast packet is multiplexed into a composite field of the header and the first segment of OPUS voice data; Subsequent voice data is transmitted cyclically at fixed time intervals on a preset channel sequence via the AUX_ADV_IND auxiliary packet to avoid 2M PHY adjacent channel interference.
3. The AI speech extension method based on BLE network according to claim 2, characterized in that: During the initial network configuration phase, the user interaction node and the relay node obtain the byte root key from the AI interaction node through a secure pairing process, and derive a session key based on the root key, which is used for MIC verification and optional encryption of the custom BLE data packets in subsequent communications. After receiving a voice data packet, the AI interaction node verifies whether its sequence number is monotonically increasing and whether the MIC verification is successful. If either verification fails, the data packet is discarded to prevent replay attacks and man-in-the-middle tampering.
4. The AI speech extension method based on BLE network according to claim 2, characterized in that: The payload of ADV_EXT_IND is multiplexed into a composite field of the guide header and the first OPUS speech frame; The subsequent voice frames are continuously transmitted on a preset non-adjacent channel sequence via the AUX_ADV_IND auxiliary packet to form voice data; The receiving node receives complete voice data through synchronous scanning and channel frequency hopping prediction mechanisms, without relying on the message segmentation and reassembly logic of BLE Mesh; The retransmission and acknowledgment mechanism of the BLE connection layer is disabled during the voice data transmission process to maintain low latency.
5. The AI speech extension method based on BLE network according to claim 4, characterized in that: The relay node performs transparent forwarding only based on the destination address field in the custom BLE data packet.
6. The AI speech extension method based on BLE network according to claim 5, characterized in that: The custom BLE data packet includes a source address field SRC_ADDR, a destination address field DST_ADDR, a 16-bit sequence number SEQ_H / L, and a message integrity check code calculated based on the session key.
7. A system applying the AI speech extension method based on BLE networks as described in any one of claims 1-6, characterized in that, include: AI interaction node, which is used to receive user voice data, generate AI voice response, and return the AI voice response to the corresponding user interaction node along the original path; The user interaction node is used to collect far-field voice from users, encapsulate the voice into a custom BLE data packet in OPUS encoding format and broadcast it to the AI interaction node, and decode and play the AI voice response returned by the AI interaction node. Each relay node contains only a BLE communication module and is configured to transparently forward voice data packets according to the target address field in the custom BLE data packet, without performing voice decoding, AI processing, or audio playback. The AI interaction node, user interaction node, and relay node together constitute a half-duplex, multi-hop BLE voice pass-through network, and there is only one AI interaction node in the entire network.
8. The AI speech extension method based on BLE network according to claim 7, characterized in that, It also includes: The AI interaction node includes a local or cloud-based AI voice processing engine. The user interaction node includes a microphone, a speaker, and a BLE communication module.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-8.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-8.