Sensor-based device for real-time acoustic condition analysis in vehicle control architectures

Abstract

Hardware-based device for real-time acoustic condition monitoring in software-defined vehicle systems, comprehensive at least one acoustic sensor unit for detecting structure-borne and airborne sound signals from at least one vehicle component, a signal preprocessing unit with analog-to-digital converter and fixed filter architecture for frequency selection and noise reduction, an AI processing unit with a dedicated accelerator processor for executing a trained neural network for state classification, a volatile and a non-volatile memory for storing model parameters, reference patterns and system configuration data, an internal, access-controlled bus architecture for connecting the aforementioned units, as well as a communication interface for transmitting a deterministically generated state parameter to a vehicle control architecture, where the signal preprocessing unit and the AI ​​processing unit are integrated on a common physical support structure and form a closed, real-time capable processing chain.

Description

Technical field

[0001] The invention relates to the field of vehicle condition monitoring systems. In particular, the invention relates to a hardware-based device for real-time acoustic condition monitoring in software-defined vehicle systems using integrated signal preprocessing, AI-based classification, and hardware-isolated processing units. The invention lies at the interface of vehicle mechatronics, embedded real-time systems, acoustic sensor technology, and accelerated AI hardware. State of the art

[0002] Modern vehicle architectures employ condition monitoring systems to detect mechanical, electrical, or structural deviations at an early stage. Known solutions capture acoustic signals using microphones or structure-borne sound sensors and transmit these to central control units or external computing units for further analysis.

[0003] Signal processing is typically software-based and performed on general-purpose microprocessors. AI-supported analysis is often implemented in virtualized environments or central vehicle computers.

[0004] These architectures have several technical disadvantages. Signal processing and classification are not fully hardware-integrated, resulting in latency. Furthermore, a physically isolated execution environment for safety-critical AI models is lacking. A closed, real-time processing chain with integrated integrity monitoring is not disclosed in the prior art.

[0005] Furthermore, no device is known in which acoustic signal preprocessing, AI accelerator, access control and integrity check are combined on a common physical support structure to form a tamper-proof unit. Object of the invention

[0006] The invention is based on the objective of providing a hardware-based device that enables real-time acoustic condition monitoring in software-defined vehicle systems with deterministic processing and increased system integrity.

[0007] In particular, an integrated architecture should be created in which acoustic sensor data acquisition, signal preprocessing, AI-based condition classification and integrity monitoring are performed within a closed, hardware-isolated processing chain.

[0008] Furthermore, the device should feature tamper-proof storage of model parameters and an access-controlled internal bus architecture. Summary of the invention

[0009] The invention solves the problem by means of a hardware-based device for real-time acoustic condition monitoring in software-defined vehicle systems.

[0010] The device includes an acoustic sensor unit for detecting structure-borne and airborne sound signals, a signal preprocessing unit with analog-to-digital converter and fixed filter architecture, an AI processing unit with dedicated accelerator processor for executing a trained neural network, and volatile and non-volatile memory for storing model parameters and system configuration data.

[0011] The aforementioned units are connected via an access-controlled internal bus architecture and integrated on a common physical support structure.

[0012] Furthermore, the device includes an integrity monitoring unit with hardware-based signature verification. If a defined integrity criterion is not met, a locking unit activates a secure system state.

[0013] The device provides a deterministically generated state parameter via a communication interface and forms a closed, real-time capable and tamper-proof processing chain within a software-defined vehicle architecture. Detailed description of the invention

[0014] The invention relates to a hardware-based device for real-time acoustic condition monitoring in software-defined vehicle systems. The device comprises at least one acoustic sensor unit for detecting structure-borne and airborne sound signals, a signal preprocessing unit with an analog-to-digital converter and a fixed filter architecture, and an AI processing unit with a dedicated accelerator processor for executing a trained neural network for condition classification. A volatile and a non-volatile memory serve to store model parameters and system configuration data. The aforementioned units are connected via an access-controlled internal bus architecture and integrated on a common physical substrate.Furthermore, an integrity monitoring unit with hardware-based signature verification is provided, which activates a locking unit if a defined integrity criterion is not met. The device provides a deterministically generated status parameter via a communication interface and ensures tamper-proof, real-time processing of acoustic signals within a software-defined vehicle architecture.

[0015] The invention relates to a hardware-based device for real-time acoustic condition monitoring in software-defined vehicle systems. The device is designed as an integrated, physically closed system unit and serves for the continuous acquisition, processing, and classification of acoustic signals from vehicle components.

[0016] The device comprises at least one acoustic sensor unit for detecting structure-borne and airborne sound signals. The sensor unit is connected to a signal preprocessing unit. The signal preprocessing unit includes an analog-to-digital converter for digitizing the detected signals and a fixed filter architecture. The filter architecture comprises hardware-implemented bandpass and noise reduction filters for frequency selection and signal stabilization.

[0017] The digitized and filtered signals are forwarded to an AI processing unit. This unit comprises a dedicated accelerator processor with a parallelized matrix processing unit and a hard-wired vector operation unit. State classification is performed by executing a trained neural network. The neural network execution is implemented within a hardware-isolated environment, thus preventing external access to internal model parameters.

[0018] For storing model parameters, reference patterns, and system configuration data, volatile and non-volatile memory are provided. Access to these memory areas is exclusively controlled by hardware-implemented access control within an internal bus architecture. This internal bus architecture is access-controlled and prevents unauthorized memory access between the individual functional units.

[0019] The sensor unit, the signal preprocessing unit, and the AI ​​processing unit are integrated on a common physical substrate. This integration forms a closed, real-time capable processing chain with deterministic signal processing within predefined time limits.

[0020] Furthermore, the device includes an integrity monitoring unit with hardware-based signature verification. Any change to model parameters or system configuration data triggers a cryptographically verified authentication check. If a defined integrity criterion is not met, a locking unit activates a secure system state. In this state, the state monitoring function is deactivated and further signal processing is prevented.

[0021] The device provides a deterministically generated state parameter via a communication interface. This state parameter represents the result of the acoustic analysis classified by the neural network and can be used within a software-defined vehicle architecture for further control or diagnostic processing.

[0022] The physical integration of all functional units, the hardware-isolated AI execution, and the integrated integrity monitoring result in a tamper-proof and real-time capable device for acoustic condition monitoring in software-defined vehicle systems.

Claims

[1] Hardware-based device for real-time acoustic condition monitoring in software-defined vehicle systems, comprehensive at least one acoustic sensor unit for detecting structure-borne and airborne sound signals from at least one vehicle component, a signal preprocessing unit with analog-to-digital converter and fixed filter architecture for frequency selection and noise reduction, an AI processing unit with a dedicated accelerator processor for executing a trained neural network for state classification, a volatile and a non-volatile memory for storing model parameters, reference patterns and system configuration data, an internal, access-controlled bus architecture for connecting the aforementioned units, as well as a communication interface for transmitting a deterministically generated state parameter to a vehicle control architecture, where the signal preprocessing unit and the AI ​​processing unit are integrated on a common physical support structure and form a closed, real-time capable processing chain. [2] Device according to claim 1, characterized by , that the AI ​​processing unit comprises a parallelized matrix processing unit with a hard-wired vector operation unit and The state classification takes place within a hardware-isolated execution environment. where access to model parameters is only permitted via a hardware-implemented access control. [3] Device according to claim 1 or 2, characterized by that the device also includes an integrity monitoring unit with hardware-based signature verification includes where any change to model parameters or system configuration data triggers a cryptographically verified authentication check, and if a defined integrity criterion is not met, a locking unit deactivates the state monitoring function and activates a secure system state.