Low-altitude capture unmanned aerial vehicle system and method based on voiceprint recognition and collaborative decision-making

Abstract

The application discloses a low-altitude capture unmanned aerial vehicle system and method based on voiceprint recognition and collaborative decision-making, and belongs to the technical field of unmanned aerial vehicle monitoring and countermeasures. The system comprises a distributed acoustic sensing network, an acoustic fingerprint recognition and intelligent analysis center, a hierarchical countermeasure response unit and a data storage and collaboration module. The system collects signals through the acoustic sensing network, identifies, locates and threat assesses based on a pre-built acoustic fingerprint library by the analysis center, and starts early warning, interference or physical interception and other hierarchical countermeasures according to the threat level. At the same time, the system stores events in a credible manner through a block chain, and collaborates data with a city low-altitude management system. The application realizes all-weather, high-precision and intelligent control and disposal of the "black flight" unmanned aerial vehicle, effectively solves the problems of insufficient detection capability, large side effects of countermeasures, data fragmentation and lack of collaboration of the prior art in the complex environment of the city, and is especially suitable for low-altitude approach defense in sensitive areas of the city.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) monitoring and countermeasure technology, and in particular to a low-altitude UAV capture system and method based on voiceprint recognition and collaborative decision-making. Background Technology

[0002] With the rapid popularization of drone technology in urban scenarios such as logistics delivery, city inspection, and cultural and sports entertainment, the problems of disorderly flight, such as "black flights" and "random flights," are becoming increasingly prominent. Especially in the approach defense of sensitive areas such as airports, government facilities, and large-scale event venues, higher demands are placed on the reliable detection and effective countermeasures of drones. The urban low-altitude environment is characterized by dense high-rise buildings, complex electromagnetic signals, and diverse background noise (such as traffic noise and environmental noise), which further increases the difficulty of technical implementation.

[0003] Currently, mainstream anti-drone technologies still have several limitations in urban applications. Traditional drone intrusion detection systems typically employ radar warnings and broad-spectrum radio jamming systems. While detecting and jamming drone intrusions, these systems also interfere with the electromagnetic environment of the surrounding area. Furthermore, they are highly susceptible to detection by drones equipped with radar countermeasures, leading to direct attacks from the target drones. In addition, other commonly used technologies also have significant shortcomings: 1. Radio detection technology: This technology relies on capturing communication signals between the drone and its remote controller. It is ineffective against drones that fly autonomously using pre-programmed methods or maintain radio silence. Furthermore, the dense concentration of civilian communication signals such as Wi-Fi and Bluetooth in urban environments easily causes interference, leading to a higher false alarm rate.

[0004] 2. Radar Detection Technology: Traditional radar has limited detection capabilities against low-speed, small-radar-cross-section "low, slow, and small" UAV targets. In urban canyon environments, radar beams are prone to multipath reflections, creating detection blind spots and false alarms. Furthermore, these systems are typically expensive and lack flexibility in deployment and mobility.

[0005] 3. Photoelectric recognition technology: This technology heavily relies on optical and infrared sensors, and its performance degrades significantly in adverse weather conditions such as nighttime, rain, snow, and fog. Its field of view cannot penetrate obstacles such as buildings, and real-time analysis of video streams requires substantial computing resources, making it difficult to achieve wide-area, all-weather coverage.

[0006] 4. Problem of limited countermeasures: Many existing systems employ a single countermeasure strategy, such as wide-area, full-band electromagnetic suppression. While this method can interfere with illegal drones, it is highly likely to cause collateral damage to legitimate wireless communications and electronic equipment in the vicinity. Furthermore, it lacks the ability to differentiate and tiered responses based on threat level, airspace attributes, and drone type, which can easily lead to insufficient countermeasures or overreactions.

[0007] 5. Data fragmentation and lack of collaboration: Data generated during detection, identification, and handling are often stored in isolation, lacking a unified and reliable data chain, which is detrimental to post-event auditing, incident tracing, and liability determination. Furthermore, existing systems generally fail to achieve deep data interaction and policy collaboration with city-level low-altitude traffic management systems, creating "information silos."

[0008] Acoustic detection technology offers unique advantages such as passive reception, immunity to electromagnetic interference, and all-weather operation, providing a new solution to the aforementioned problems, and is particularly suitable for approach defense against intruding drones. However, existing acoustic-based solutions generally have significant shortcomings: insufficient recognition accuracy and reliability in complex urban noise environments; a lack of intelligent and precise closed-loop linkage between recognition results and subsequent countermeasures; and failure to integrate into the urban airspace management ecosystem. Therefore, the industry urgently needs a comprehensive solution that can deeply integrate accurate acoustic recognition, intelligent decision-making and countermeasures, reliable data storage, and systematic collaborative management. Summary of the Invention

[0009] To overcome the shortcomings of existing technologies, this application provides a low-altitude drone capture system and method based on voiceprint recognition and collaborative decision-making. This system integrates key technologies such as distributed acoustic sensing, intelligent voiceprint recognition, hierarchical precise countermeasures, blockchain-based trusted evidence storage, and urban air traffic control collaboration to construct a closed-loop solution encompassing "detection-identification-decision-countermeasures-evidence storage-collaboration," effectively achieving all-weather, high-precision, and intelligent control and disposal of urban low-altitude "black flight" drones.

[0010] To achieve the above objectives, this application provides the following solution: A low-altitude capture UAV system and method based on voiceprint recognition and collaborative decision-making, characterized in that it includes: Distributed acoustic sensor networks are used to collect raw acoustic signals in the monitored airspace. The acoustic fingerprint recognition and intelligent analysis center is connected to the distributed acoustic sensor network for processing the raw acoustic signal, extracting acoustic features, identifying and locating the target based on the pre-built acoustic fingerprint database, and outputting target information including the drone model, location, speed and threat level. The hierarchical countermeasure response unit is communicatively connected to the acoustic fingerprint recognition and intelligent analysis center. It is used to receive the target information and select and activate the corresponding countermeasure strategy based on the drone model, location, speed and threat level contained therein. The data storage and collaboration module is communicatively connected to the acoustic fingerprint recognition and intelligent analysis center and the hierarchical countermeasure response unit. It is used to record system event data, perform credible storage, and exchange data with the low-altitude management system.

[0011] Furthermore, the distributed acoustic sensor network includes multiple acoustic sensor nodes deployed in an array, each node having a built-in microphone array module; the nodes are spaced 0.5m to 2m apart, to achieve acoustic signal coverage and high-precision time difference positioning in a spherical airspace with a radius of not less than 500m.

[0012] Furthermore, the acoustic fingerprint recognition and intelligent analysis center includes: a preprocessing unit for noise reduction, framing, and gain control of the acoustic signal; a feature extraction unit for extracting time-frequency domain features of the signal, the features including at least Mel frequency cepstral coefficients and wavelet packet energy features; and a matching and identification unit for comparing the extracted features with the acoustic fingerprint database to identify the UAV model and calculating the UAV position based on the signal arrival time difference of multiple sensor nodes.

[0013] Furthermore, the acoustic fingerprint database stores 1200 preset acoustic feature templates for different types of UAVs under different flight states; the matching recognition and positioning unit adopts a hybrid recognition strategy that combines dynamic time warping algorithm and deep learning classification model.

[0014] Furthermore, the tiered countermeasure response unit includes an early warning subunit, used to send alarm information to a preset management entity after identifying an unauthorized or suspicious drone; at least one physical interception subunit, used to capture the target drone or force it to land safely; and a strategy engine, used to sequentially schedule the early warning and physical interception subunits according to the threat level and target location information, wherein the execution of all countermeasures is primarily constrained by not generating abnormal noise, strong light, or electromagnetic radiation that can be perceived by the public.

[0015] Furthermore, the alarm information sent by the early warning subunit includes at least the target location, model, speed, threat level, and suggested response measures, and is sent to the monitoring terminal via a data link or secure network; the physical interception subunit includes one or more of the following: net-trapping drones, laser interception systems, or parachute decoy launchers.

[0016] Furthermore, the strategy engine is configured to execute the following tiered countermeasure logic: when an unauthorized drone is detected, the early warning subunit is activated first; if the target continues to linger or moves toward a sensitive area and the threat level reaches medium or high, the physical interception subunit is scheduled to perform an interception task.

[0017] Furthermore, the data storage and collaboration module includes a blockchain storage unit, used to store the hash values ​​of key events, including early warning events, identification trajectories, and countermeasures, in the consortium blockchain; and an urban low-altitude management interface unit, used to report the identification, early warning, and handling information of drones to the urban low-altitude management system in real time, and to obtain legitimate flight plans from the urban low-altitude management system for identity verification.

[0018] A method for detecting low-altitude anti-drone drones in urban areas based on voiceprint recognition and collaborative decision-making is characterized by the following steps: S1: Signal acquisition, synchronously acquiring low-altitude acoustic signals through a distributed acoustic sensor network; S2: Intelligent identification, which preprocesses and extracts features from the collected signals and matches them with the acoustic fingerprint database to identify the drone model, location, speed and assess its threat level; S3: Early warning response. Upon identifying an unauthorized or suspicious target, the early warning procedure is activated first to send an alarm to the management. S4: Tiered countermeasures: Based on the identified threat level, select and trigger corresponding countermeasures, including early warning subunits, physical interception subunits, and combinations thereof to execute countermeasure actions; S5: Evidence storage collaboration, which synchronizes key event data throughout the identification and countermeasure process for blockchain evidence storage and exchanges information with the city's low-altitude management system.

[0019] Furthermore, the acoustic fingerprint database also has an online incremental learning and updating mechanism; when the matching and identification unit identifies a new type of UAV target that is not archived in the database with high confidence based on the deep learning classification model, the system will trigger a composite verification process involving optical or radio-assisted sensors; after the verification is passed, the acoustic feature template of the new target and the corresponding flight status metadata will be automatically extracted, reviewed and stored in the acoustic fingerprint database, realizing the dynamic expansion of the fingerprint database.

[0020] The beneficial effects of the technical solution in this application are as follows: 1. Achieved all-weather, high-precision drone detection and identification in complex urban environments. The system employs passive acoustic detection, unaffected by electromagnetic interference, nighttime conditions, or inclement weather, overcoming the limitations of traditional radar and radio detection technologies in urban environments. Through a distributed high-sensitivity microphone array network and a database containing 1200 preset acoustic fingerprints, combined with a hybrid recognition strategy of dynamic time warping and deep learning, the system significantly improves the accuracy and model identification capability of identifying "low, slow, and small" drones in strong background noise.

[0021] 2. A tiered countermeasure system of "early warning first, precise intervention, and physical termination" has been established, completely avoiding electromagnetic interference. The system responds first with early warning, providing non-intrusive alerts via data links or high-frequency directional sound. The core countermeasure employs a combination of acoustic camouflage and physical interception: acoustic camouflage uses simulated ambient sound for concealment and deception; physical interception (netting, laser, decoys, etc.) achieves precise capture. The entire countermeasure process strictly adheres to the principle of "not generating any abnormal noise, strong light, or electromagnetic radiation that can be perceived by the public," effectively addressing threats while maximizing public environmental safety and electromagnetic compatibility, making it particularly suitable for deployment in sensitive areas such as densely populated areas and around important facilities.

[0022] 3. The system integrates a blockchain-based evidence storage unit based on a consortium blockchain, which records key event data such as drone identification, location, threat assessment, and countermeasure execution on the blockchain in real time, forming an immutable and multi-party consensus-based trusted evidence storage chain. This provides reliable technical support for post-event auditing, event tracing, and the definition of responsibilities, meeting the compliance and transparency requirements of urban public safety management.

[0023] 4. It achieves deep collaboration with the urban low-altitude airspace management system, breaking down "information silos." Through standardized interfaces, it enables two-way data exchange with the urban low-altitude airspace management system, allowing the system to obtain legal flight plans across the entire airspace in real time, greatly improving the accuracy of threat assessment. At the same time, it reports the system's monitoring, early warning, and response information in real time, facilitating the integration and coordinated scheduling of urban-level air traffic situations and promoting the systematization and intelligentization of low-altitude airspace management. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a diagram of the architecture of an urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making. Figure 2 A flowchart of acoustic fingerprint recognition and positioning for an urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making. Figure 3 A hierarchical countermeasure response logic diagram for an urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making. Figure 4 This is a flowchart of an urban low-altitude anti-drone method based on voiceprint recognition and collaborative decision-making. Figure 5This is a schematic diagram illustrating the deployment and application scenarios of an urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the specific embodiments and accompanying drawings are described in detail below. This description is intended to provide sufficient information for those skilled in the art to understand and implement this invention. The specific details and parameters described below should not be considered as limitations on the scope of protection of this invention, but are merely examples of preferred embodiments. For those skilled in the art, various modifications to parameters, algorithm implementations, or engineering optimizations made without departing from the spirit and substance of this invention fall within the scope of protection of the claims of this invention.

[0027] This system consists of four core functional modules connected by a communication network to form a collaborative whole. The distributed acoustic sensor network, acting as the front-end perception layer, is responsible for collecting raw analog acoustic signals within a designated airspace, converting them into digital signals, and uploading them. The acoustic fingerprint recognition and intelligent analysis center, acting as the core processing layer, receives digital signals from the front end, executes signal processing, feature extraction, pattern matching, and threat assessment algorithms, and outputs structured target information. The hierarchical countermeasure response unit, acting as the execution layer, receives target information and selects and drives specific countermeasure devices based on embedded decision logic. The data storage and coordination module, acting as the data and coordination layer, runs through the above three processes, responsible for recording key events, achieving data storage, and conducting bidirectional data interaction with external platforms such as the urban low-altitude airspace management system.

[0028] Each acoustic sensor node is an independent embedded system. Its hardware core includes: an array of at least four high-sensitivity MEMS microphones arranged in a specific geometry (linear or circular) for acquiring spatial sound field signals; a signal conditioning circuit board integrating an analog front-end and an analog-to-digital converter; an embedded processor for running simple local preprocessing algorithms and a network protocol stack; and a communication module. The physical deployment of the nodes needs to be optimized according to the accuracy requirements of the generalized cross-correlation time delay estimation method for sound source localization. Configuring the node spacing between 0.5m and 2m provides the optimal baseline length for the time difference of arrival-based localization algorithm in a typical urban sound speed environment of approximately 340m / s, balancing localization accuracy and construction cost. Based on a model calculating the sound pressure level attenuation with distance generated by mainstream consumer drones flying at an altitude of 50 meters, the effective acoustic detection radius of a single node is designed to be no less than 500 meters, ensuring continuous network coverage of the protected airspace.

[0029] The Acoustic Fingerprint Recognition and Intelligent Analysis Center is divided into three logical units in terms of software: preprocessing, feature extraction, and matching recognition.

[0030] The preprocessing unit performs the following steps in sequence: adaptive noise suppression, which uses a filter based on the recursive least squares algorithm to estimate and filter out steady-state background interference such as traffic noise in real time; beamforming, which uses the geometric information of the microphone array to enhance signals from specific directions; and frame windowing, which divides the continuous signal into overlapping short frames, with a preferred frame length of 25 milliseconds, a frame shift of 10 milliseconds, and applies a Hamming window to reduce spectral leakage.

[0031] The feature extraction unit extracts two types of features in parallel for each frame of signal: First, the 13th-order Mel-frequency cepstral coefficients, obtained by calculating the short-time power spectrum of the signal, passing it through a set of Mel-scale triangular filters, taking the logarithm, and then performing a discrete cosine transform, used to characterize the spectral envelope related to human auditory perception. Second, the energy features of the 4-level wavelet packet decomposition, using the Db4 wavelet basis function to perform a complete tree decomposition of the signal, calculating the normalized energy of all 16 sub-bands at the 4th level, used to characterize the fine distribution and transient characteristics of the signal in the frequency domain. These two types of features together constitute a 29-dimensional feature vector.

[0032] The matching and identification unit employs a hybrid identification strategy. For models already archived in the acoustic fingerprint database, an improved dynamic time warping algorithm is used for rapid matching. This algorithm introduces global path constraints and local distance weights to improve robustness to feature sequence deformation caused by changes in flight speed. Simultaneously, a deep learning model based on a combination of a one-dimensional convolutional neural network and a long short-term memory network is deployed as an auxiliary classifier. This model takes the aforementioned feature vector sequence as input and directly outputs the classification results for unknown or low-confidence targets. The results are then fused with those from the DTW algorithm to improve the system's adaptability to new models and complex noise environments.

[0033] The construction of the acoustic fingerprint database comprises three stages: data acquisition, feature engineering, and template management. The data acquisition stage is conducted in a semi-anechoic chamber and various typical urban outdoor environments, systematically recording 1200 different models and sizes of drones in various flight states, including hovering, constant speed movement, climbing, descending, and turning, using standard recording equipment. In the feature engineering stage, each sample undergoes the same preprocessing and feature extraction procedures as described above, generating a standardized feature template sequence. Each template is labeled with metadata such as drone model, flight state, and acquisition environment. In the template management stage, a multi-level index database is established for the feature templates according to dimensions such as model, brand, power type, and overall payload, supporting efficient retrieval and comparison. The database supports an online update mechanism. When an auxiliary classifier identifies an unidentified target with high confidence, and this is verified by other sensors, its feature template can be added to the database after review, enabling continuous evolution of the fingerprint database.

[0034] The acoustic fingerprint database possesses online incremental learning and dynamic update capabilities, allowing it to continuously evolve with the iteration of drone models. Its update mechanism is as follows: When a deep learning classification model identifies a target with high confidence, but the dynamic time warping algorithm has insufficient matching degree in the existing database, the system marks it as a "suspected new model," automatically triggering a composite verification process involving optical and radio sensors to eliminate interference and confirm the target's identity. After successful verification, the system extracts high-quality acoustic signals from the target, performs the same preprocessing and feature extraction process as during database creation, and generates a standard feature template sequence. This template, along with relevant metadata, enters an automated review queue. Once approved, it is temporarily stored as a candidate template and officially added to the database after final confirmation by the administrator. After the fingerprint database is updated, the system can initiate background incremental training tasks, using new data to fine-tune the deep learning model, continuously improving recognition accuracy and generalization ability. This mechanism significantly enhances the system's adaptability to new threats and extends the technology lifecycle.

[0035] The tiered countermeasure response unit physically integrates control units for two types of countermeasure devices: an early warning subunit and a physical interception subunit. The early warning subunit establishes a secure data link (such as a 4G / 5G private network or fiber optic cable) with the monitoring center's command platform and mobile terminals via the system's main control unit's data interface. When the system identifies an unauthorized or suspicious drone, the early warning subunit automatically generates structured alarm information, including the target's real-time location, model, speed, heading, threat level, and suggested response measures, and immediately pushes it to the pre-set monitoring terminal via an encrypted data link.

[0036] The physical interception subunit can connect to and deploy various precision interception equipment, including but not limited to: net-capture drones, laser interception systems, and parachute decoy launchers. All physical interception methods aim to capture or safely force a landing, avoiding collateral damage to personnel and property on the ground.

[0037] The strategy engine is the intelligent decision-making core of the hierarchical countermeasure response unit, embedded in the system's main control unit. Based on target information output by the acoustic fingerprint recognition and intelligent analysis center, including model, location, speed, and threat level, and combined with a predefined rule base, it executes the following hierarchical response logic: 1. Early warning priority. When an unauthorized drone is detected, regardless of the threat level, the policy engine first activates the early warning subunit to send an alarm message to the management.

[0038] 2. Physical Interception Termination. If the threat level reaches "Medium" or "High," or if the target continues to pose a threat, the policy engine will schedule the physical interception subunit. Based on the target's location, speed, and airspace environment, the most suitable interception equipment will be selected, and control commands including target coordinates, interception path, and equipment parameters will be generated and executed via CAN bus or private wireless network.

[0039] The scheduling and execution of all countermeasures are subject to the primary constraint of "not generating any abnormal noise, strong light, or electromagnetic radiation that can be perceived by the public," in order to ensure the effectiveness and environmental friendliness of the countermeasures.

[0040] The data storage and collaboration module includes a blockchain storage unit. This unit is implemented using the Hyperledger Fabric consortium blockchain framework. The consensus nodes of the consortium blockchain are deployed and maintained by authoritative institutions such as airspace regulatory authorities, public security organs, and system operators. The storage unit is automatically triggered after the system completes a full event handling process from identification to countermeasures. It serializes the standardized event data, including: event ID, UTC timestamp, target latitude and longitude coordinates, identified model and confidence level, threat level, countermeasure code taken, and handling result status code, and calculates its SHA-256 hash value. This hash value, serving as the "fingerprint" of the transaction data subject, is packaged together with the digital signature of the submitting node into a transaction proposal and submitted to the consortium blockchain network. After each consensus node verifies the legality of the transaction through a pre-built smart contract, it sorts and packages the transactions into a new block according to a predetermined consensus mechanism. The block is then broadcast and appended to the ledger copies of each node, forming an immutable storage chain.

[0041] The data storage and collaboration module also includes an urban low-altitude airspace management interface unit. This interface maintains two continuous data stream sessions based on the MQTT message queue protocol: the uplink data stream session is responsible for encapsulating the system's output into JSON or Protobuf format messages according to relevant industry standard protocols and pushing them in real time to the designated data receiving endpoint of the urban low-altitude airspace management system. The content mainly includes the real-time trajectory points, attributes, warning records, and handling status of illegal intrusion targets. The downlink data stream session obtains details of all approved legal drone flight plans within the city from the urban low-altitude airspace management system through subscription or polling, including the aircraft's unique identifier, planned takeoff / landing time, approved 3D waypoints, and electronic fence information. After receiving a detected target, the system's threat assessment module first performs real-time matching and verification with the legal flight plan database obtained from the downlink. If no matching legal plan is found, the target is immediately marked as "unauthorized," and its weight coefficient in the threat assessment model is significantly increased, thereby triggering the warning and subsequent response process.

[0042] Based on the specific implementation of the above modules, the automated execution flow of the drone capture method defined in this application is as follows: Step S1: Signal Acquisition and Upload. Each node in the distributed acoustic sensor network synchronously acquires acoustic signals at a sampling rate of 48kHz. After local pre-filtering, the signals are precisely timestamped using a precision time protocol and then streamed to the central server or edge computing nodes via 5G or fiber optic networks.

[0043] Step S2: Intelligent Identification and Assessment. The central server invokes the preprocessing, feature extraction, and matching identification units to suppress noise and extract features from the signal, and queries the acoustic fingerprint database for matching identification. Simultaneously, it calculates the target location based on multi-node time delay information. The identification and positioning results, along with authorization information obtained from the city's low-altitude management system interface, are fed into the threat assessment model to generate a comprehensive flight track containing the target ID, model, location, speed, heading, and threat level.

[0044] Step S3: Warning Response. Once the integrated track is marked as "unauthorized" or "suspicious," the system immediately initiates the warning procedure, and the warning subunit sends an alarm message containing target details to the monitoring center.

[0045] Step S4: Tiered Countermeasure Decision and Execution. The strategy engine of the tiered countermeasure response unit receives the integrated flight path generated in Step S2 and the early warning feedback from Step S3. Based on the dynamic changes in the target threat level, the geofenced area it is located in, and the preset response rule table, the engine decides whether and how to activate the physical interception subunit. For example, for low-threat targets that have not entered the core area, it may only continuously issue warnings and record them; for targets that still move towards the no-fly zone after issuing warnings, and are clearly medium- or high-threat targets that have intruded into the core area, physical interception equipment is directly dispatched to capture them.

[0046] Step S5: Data Storage and Collaboration. Simultaneously with steps S3 and S4, the blockchain storage unit is invoked to generate and upload the key data of this event to the blockchain as evidence. The city's low-altitude airspace management interface unit is also triggered, pushing a full report of this event to the uplink, completing the collaborative loop with the city-level management platform.

[0047] Example 1: Collaborative response in the scenario of perimeter protection for key facilities; This example uses the perimeter protection of a key infrastructure as an application scenario to illustrate the collaborative workflow of the various modules of this system.

[0048] The system was first deployed around the perimeter of the facility. Nodes of the distributed acoustic sensor network were arranged in an equilateral triangular grid with a 1.5-meter spacing. Each node incorporates a quaternary MEMS microphone linear array, a signal conditioning unit, and a 5G communication module. The node deployment height was specifically optimized to ensure effective acoustic coverage of a spherical airspace with a radius of 500 meters and a height of 300 meters or less.

[0049] At a certain moment, nodes A, B, and C nearly synchronously detect an acoustic signal containing significant high-frequency harmonic components. Each node then performs local pre-filtering on the raw signal and adds a precise timestamp synchronized via the Precision Time Protocol (PTP). Subsequently, the timestamped multi-channel audio data stream is uploaded in real-time via a 5G network to an acoustic fingerprinting and intelligent analysis center deployed at the edge cloud.

[0050] After receiving the signal, the analysis center first performs an adaptive beamforming algorithm on the three received audio streams using a preprocessing unit. Based on the known geometric parameters of the node array, the signal component from the northeast direction is enhanced while effectively suppressing background noise from other directions. Next, the signal is framed and subjected to a Hamming window for subsequent analysis. The feature extraction unit then calculates the 13th-order Mel-frequency cepstral coefficients and the energy features of the four-layer wavelet packet decomposition in parallel for each frame, combining these two types of features into a 29-dimensional feature vector.

[0051] The matching and identification process employs a hybrid strategy. An improved dynamic time warping algorithm rapidly compares the extracted feature sequence with 1200 pre-set acoustic templates for different UAV models in an acoustic fingerprint database, initially determining that the current target has an 89% similarity to the template of a "certain commercial quadcopter UAV" in level flight. Simultaneously, a parallel-running one-dimensional convolutional neural network-long short-term memory deep learning model analyzes the same feature sequence, and the output classification probability supports this judgment. After decision-level fusion, the system ultimately confirms the target as the UAV model and provides an 89% confidence level. During identification, the positioning unit utilizes the precise time difference of the same sound signal arriving at nodes A, B, and C, combined with known node coordinates and a sound speed model, to calculate that the target is currently located 500 meters northeast of the facility at an altitude of 500 meters, and estimates its speed at 8 meters per second, with its heading towards the facility's core area.

[0052] The threat assessment module integrates multiple pieces of information for judgment. This module receives results from the identification and location unit. Simultaneously, the urban low-altitude airspace management interface unit in the data storage and collaboration module queries and confirms in real-time from the urban low-altitude airspace management system via downlink data stream: there are no approved legal flight plans in this airspace during this period. Based on a preset model, the assessment module comprehensively considers key factors such as the target's commercial-grade model, large potential payload capacity, flight trajectory continuously moving towards the core area, high speed, and lack of registration with the low-altitude air traffic management system, classifying the target's threat level as "high."

[0053] Based on the "high threat" determination, the strategy engine of the tiered countermeasure response unit is triggered, executing the following sequence: the early warning subunit immediately sends a red alert to the facility security command center, including all parameters of the target; at the same time, it activates one or more of the friendly defense drone systems near the target's location, including net-capture drones, attack drones, and laser interception systems, to quickly deal with the situation.

[0054] Throughout the entire handling process and afterward, the data storage and collaboration module operated synchronously. Its blockchain storage unit was automatically triggered, serializing standardized data for the event, including event ID, UTC timestamp, target model / coordinates / speed, threat level determination criteria, and initiated countermeasure codes, calculating its SHA-256 hash value, and packaging it into a transaction with the system operator's digital signature. This transaction was submitted to a consortium blockchain network jointly maintained by airspace regulatory authorities, public security organs, and the system operator. After verification through the consensus mechanism, it was included in a new block, forming an immutable storage record. On the other hand, the urban low-altitude management interface unit, through uplink data flow, pushed the real-time trajectory, attributes, and handling status of this "unauthorized drone intrusion" event to the command and control platform of the urban low-altitude management system according to a standard protocol format, thereby updating the global air situation map and achieving airspace collaboration.

[0055] Example 2: Low-perception handling in large-scale event security scenarios; This example uses the security work during a concert held at a large city stadium as a scenario to demonstrate the system's ability to handle low-threat targets in a refined and covert manner.

[0056] The system deployed acoustic sensor nodes on the stadium roof and surrounding light poles to construct a three-dimensional monitoring network. During the performance, the system identified a "micro recreational drone of a certain model" hovering and loitering approximately 180 meters west of the stadium at a height of 100 meters, with a confidence level of 96%. By querying the city's low-altitude traffic system, it was confirmed that there were no reported plans in that airspace at that time. Considering its miniature size and loitering behavior, the system determined its threat level to be "low," but imposed the constraint of "avoiding causing public panic."

[0057] Based on this, the strategy engine invokes refined handling rules: a yellow alert is sent to the mobile terminal of the on-site security manager, indicating "micro drone, low threat, observation recommended". If further intrusion is detected by the drone, the friendly defense drone system near the target location is activated, including one or more measures such as net-capture drones, attack drones, and laser interception systems for rapid handling.

[0058] Simultaneously, the system's data storage and collaboration module completed task recording. The blockchain storage unit encrypted and uploaded the complete log of this low-collateral damage handling event, including the triggering cause, the acoustic camouflage and interference parameters used, and the final handling result, to the blockchain, forming an immutable storage record. The urban low-altitude management interface unit also reported this event as a successful "low-altitude minor incident handling case" to the urban low-altitude management system, providing a valuable practical case for the city-level low-altitude management database and realizing closed-loop handling and information collaboration.

[0059] The above embodiments illustrate in detail the working principle, core algorithm, and control strategy of the present invention. It should be noted that the above description includes specific details and algorithmic ideas, but should not be construed as limiting the present invention. For those skilled in the art, any modifications made to parameters, specific algorithm implementations, or engineering optimizations without departing from the core concept of the present invention—namely, based on air-ground spatial coupling analysis and through the synchronous execution of air-ground cooperative control commands by modified distributed beacon nodes—should fall within the scope of protection claimed by the present invention.

Claims

1. An urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making, characterized in that, include: Distributed acoustic sensor networks are used to collect raw acoustic signals in the monitored airspace. The acoustic fingerprint recognition and intelligent analysis center is connected to the distributed acoustic sensor network for processing the raw acoustic signal, extracting acoustic features, identifying and locating the target based on the pre-built acoustic fingerprint database, and outputting target information including the drone model, location, speed and threat level. The hierarchical countermeasure response unit is communicatively connected to the acoustic fingerprint recognition and intelligent analysis center. It is used to receive the target information and select and activate the corresponding countermeasure strategy based on the drone model, location, speed and threat level contained therein. The data storage and collaboration module is communicatively connected to the acoustic fingerprint recognition and intelligent analysis center and the hierarchical countermeasure response unit. It is used to record system event data, perform credible storage, and exchange data with the low-altitude management system.

2. The urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making according to claim 1, characterized in that, The distributed acoustic sensor network includes multiple acoustic sensor nodes arranged in an array, each node having a built-in microphone array module; the nodes are spaced 0.5m to 2m apart, which is used to achieve acoustic signal coverage and high-precision time difference positioning in a spherical airspace with a radius of not less than 500m.

3. The urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making according to claim 2, characterized in that, The acoustic fingerprint recognition and intelligent analysis center includes: a preprocessing unit for noise reduction, framing, and gain control of acoustic signals; a feature extraction unit for extracting time-frequency domain features of the signal, the features including at least Mel frequency cepstral coefficients and wavelet packet energy features; and a matching and identification unit for comparing the extracted features with the acoustic fingerprint database to identify the UAV model and calculating the UAV position based on the time difference of arrival of signals from multiple sensor nodes.

4. The urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making according to claim 3, characterized in that, The acoustic fingerprint database stores 1200 preset acoustic feature templates of different drone models under different flight conditions; the matching recognition and positioning unit adopts a hybrid recognition strategy that combines dynamic time warping algorithm and deep learning classification model.

5. The urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making according to claim 1, characterized in that, The tiered countermeasure response unit includes an early warning subunit, which prioritizes sending alarm information to a pre-defined management entity upon identifying an unauthorized or suspicious drone; at least one physical interception subunit, which performs the capture of the target drone or forces it to land safely; and a strategy engine, which sequentially schedules the early warning and physical interception subunits according to the threat level and target location information, wherein the execution of all countermeasures is primarily constrained by the requirement that no abnormal noise, strong light, or electromagnetic radiation perceptible to the public be generated.

6. The urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making according to claim 5, characterized in that, The warning information sent by the early warning subunit includes at least the target location, model, speed, threat level, and suggested response measures, and is sent to the monitoring terminal via a data link or security network; the physical interception subunit includes one or more of the following: net-capture drones, laser interception systems, or parachute decoy launchers.

7. The urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making according to claim 6, characterized in that, The strategy engine is configured to execute the following tiered countermeasure logic: when an unauthorized drone is detected, the early warning subunit is activated first; if the target continues to linger or moves toward a sensitive area and the threat level reaches medium or high, the physical interception subunit is scheduled to perform an interception task.

8. The urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making according to claim 1, characterized in that, The data storage and collaboration module includes a blockchain storage unit, which stores the hash values ​​of key events, including early warning events, identification trajectories, and countermeasures, in the consortium blockchain; and an urban low-altitude management interface unit, which reports the identification, early warning, and handling information of drones to the urban low-altitude management system in real time, and obtains legal flight plans from the urban low-altitude management system for identity verification.

9. A method for capturing low-altitude unmanned aerial vehicles based on voiceprint recognition and collaborative decision-making, characterized in that, Includes the following steps: S1: Signal acquisition, synchronously acquiring low-altitude acoustic signals through a distributed acoustic sensor network; S2: Intelligent identification, which preprocesses and extracts features from the collected signals and matches them with the acoustic fingerprint database to identify the drone model, location, speed and assess its threat level; S3: Early warning response. Upon identifying an unauthorized or suspicious target, the early warning procedure is activated first to send an alarm to the management. S4: Tiered countermeasures: Based on the identified threat level, select and trigger corresponding countermeasures, including early warning subunits, physical interception subunits, and combinations thereof to execute countermeasure actions; S5: Evidence storage collaboration, which synchronizes key event data throughout the identification and countermeasure process for blockchain evidence storage and exchanges information with the city's low-altitude management system.

10. The urban low-altitude anti-drone system based on voiceprint recognition and collaborative decision-making according to claim 4, characterized in that, The acoustic fingerprint database also has an online incremental learning and updating mechanism; when the matching and identification unit identifies a new type of UAV target that is not archived in the database with high confidence based on the deep learning classification model, the system will trigger a composite verification process involving optical or radio-assisted sensors. Once verified, the acoustic feature template of the new target and the corresponding flight status metadata will be automatically extracted, reviewed, and stored in the acoustic fingerprint database, thereby enabling dynamic expansion of the fingerprint database.

Images