CN-121725784-B - BCM module cooperative control method based on voice instruction recognition under vehicle-mounted high-noise environment

CN121725784BCN 121725784 BCN121725784 BCN 121725784BCN-121725784-B

Abstract

The invention relates to the technical field of artificial intelligence and discloses a BCM module cooperative control method based on voice instruction recognition in a vehicle-mounted high-noise environment, which comprises the following steps of collecting original voice instruction data and vehicle state parameter data in the vehicle-mounted high-noise environment; the noise-removed voice command data is generated by preprocessing the voice signal in the noise environment based on the original voice command data, engine noise and wind noise steady background noise are filtered through multi-level noise suppression processing combining self-adaptive filtering and spectral subtraction, and residual noise is eliminated aiming at sudden impact noise, so that the definition of the voice signal is improved. Meanwhile, through a voice activity detection algorithm, and the detection threshold value of the voice activity detection algorithm is dynamically adjusted according to the vehicle state parameters, an effective voice segment and a noise segment can be separated, and the accuracy of voice feature extraction under a complex time-varying noise environment is ensured, so that the robustness and reliability of voice instruction recognition are improved.

Inventors

Liao xiong
LIN HAIYANG
CHEN MEIXUAN

Assignees

厦门华聚物联技术有限公司

Dates

Publication Date: 20260512
Application Date: 20260210

Claims (9)

1. The BCM module cooperative control method based on voice command recognition in the vehicle-mounted high-noise environment is characterized by comprising the following steps: s1, acquiring original voice instruction data and vehicle state parameter data in a vehicle-mounted high-noise environment; s2, preprocessing a voice signal in a noise environment based on the original voice instruction data to generate denoising voice instruction data; S3, voice instruction feature extraction processing is carried out according to the denoising voice instruction data, and voice instruction feature vector data are generated; S4, carrying out voice command recognition model matching processing based on the voice command feature vector data to generate preliminary voice command recognition result data; s5, carrying out BCM cooperative control strategy analysis by combining the vehicle state parameter data to generate BCM control instruction candidate set data, wherein the method comprises the following steps of: s51, determining an executable BCM control instruction type according to the vehicle speed and the engine speed selected from the vehicle state parameter data; S52, in the executable BCM control instruction type range, generating a BCM control instruction candidate set based on instruction frequency in historical control data and risk level in a safety rule; s53, carrying out priority ordering on candidate instructions by a fuzzy logic algorithm in combination with historical selection data in a user preference model to generate BCM control instruction candidate set data; S6, performing instruction consistency verification processing based on the preliminary voice instruction recognition result data and the BCM control instruction candidate set data to generate optimized voice instruction recognition data; S7, performing BCM cooperative control instruction execution processing according to the optimized voice instruction identification data to generate BCM control execution feedback data; S8, performing performance evaluation and self-adaptive adjustment based on the BCM control execution feedback data to generate dynamic parameter update data; and S9, optimizing the voice recognition model and the BCM control strategy in real time by utilizing the dynamic parameter updating data, and generating self-adaptive cooperative control model data.
2. The BCM module cooperative control method based on voice command recognition under the vehicle-mounted high-noise environment of claim 1, wherein the step of collecting original voice command data and vehicle state parameter data under the vehicle-mounted high-noise environment in S1 comprises the following steps: s11, acquiring original voice instruction data in a high-noise environment in a vehicle through a vehicle-mounted microphone array, wherein the high-noise environment comprises engine noise, wind noise and road noise; S12, acquiring vehicle state parameter data including vehicle speed, engine speed, vehicle window state and air conditioner operation parameters through a vehicle CAN bus; S13, synchronizing the time stamp of the original voice instruction data and the vehicle state parameter data, and storing the time stamp to an on-board embedded database.
3. The BCM module cooperative control method based on voice command recognition under the vehicle-mounted high-noise environment as set forth in claim 2, wherein the step of preprocessing the voice signal under the noise environment in S2 comprises the following steps: s21, performing background noise suppression processing on the original voice instruction data by adopting an adaptive filter to generate preliminary denoising voice data; S22, carrying out residual noise elimination on the primary denoising voice data based on spectral subtraction to generate secondary denoising voice data; S23, extracting effective voice segments through a voice activity detection algorithm, and generating denoising voice instruction data, wherein a voice activity detection threshold value is dynamically adjusted according to vehicle state parameters.
4. The BCM module cooperative control method based on voice command recognition under the vehicle-mounted high-noise environment as claimed in claim 3, wherein the voice command feature extraction processing in S3 comprises the following steps: S31, carrying out framing and windowing on the denoising voice instruction data, wherein the frame length is 25ms, and the frame shift is 10ms; s32, extracting the mel frequency cepstrum coefficient characteristic and the linear predictive coding characteristic of each frame of voice to generate a basic characteristic vector; S33, combining the context information, performing time sequence feature fusion through a cyclic neural network, and generating voice instruction feature vector data.
5. The BCM module cooperative control method based on voice command recognition under the vehicle-mounted high-noise environment as set forth in claim 4, wherein the voice command recognition model matching processing performed in S4 comprises the following steps: s41, establishing a voice instruction recognition model library, wherein the voice instruction recognition model library comprises a hidden Markov model, a deep neural network model and an end-to-end sequence model; s42, inputting the voice instruction feature vector data into the model library for parallel recognition, and generating a plurality of candidate recognition results with confidence scores; and S43, selecting an optimal candidate based on the confidence score of the candidate recognition result and the vehicle state parameter data, and generating preliminary voice instruction recognition result data.
6. The BCM module cooperative control method based on voice command recognition under the vehicle-mounted high-noise environment as set forth in claim 1, wherein the step of performing command consistency verification processing in S6 comprises the following steps: S61, carrying out semantic matching analysis on the preliminary voice instruction recognition result data and the BCM control instruction candidate set data to generate a similarity evaluation result; s62, carrying out instruction consistency verification based on the similarity evaluation result to generate verification conclusion data; And S63, directly outputting optimized voice instruction recognition data when the verification conclusion is consistent, otherwise, triggering a re-recognition flow by combining with the context information.
7. The BCM module cooperative control method based on voice command recognition in a vehicle-mounted high-noise environment according to claim 6, wherein the BCM cooperative control command execution process performed in S7 includes the steps of: S71, analyzing the optimized voice command recognition data into a control command sequence through a BCM communication interface; s72, executing the control instruction sequence and monitoring the change of the execution state of the BCM control instruction in real time to generate state log data; S73, when the state is abnormal, a backup control mechanism is started, and BCM control execution feedback data is generated.
8. The method for BCM module cooperative control based on voice command recognition in a vehicle-mounted high-noise environment according to claim 7, wherein the performance evaluation and adaptive adjustment in S8 comprises the steps of: s81, executing feedback data based on the BCM control, analyzing and generating a performance evaluation report, wherein the performance evaluation report comprises identification accuracy and response efficiency indexes; s82, dynamically optimizing internal parameters of a voice recognition model and an environmental noise adaptation threshold according to the performance evaluation report; s83, generating and uploading a transmission state parameter updating data packet to a cloud server based on the optimized parameters and the threshold value, and performing iterative learning.
9. The method for BCM module cooperative control based on voice command recognition in a vehicle-mounted high-noise environment according to claim 8, wherein the step of optimizing the voice recognition model and the BCM control strategy in real time in S9 comprises the steps of: S91, calling the dynamic parameter updating data packet, and performing incremental training and optimization on the voice recognition model through a machine learning algorithm; s92, placing the optimized model in a virtual BCM control scene for simulation verification, and generating a strategy effectiveness verification result; s93, fusing the verification result and the current model data, updating and outputting the self-adaptive cooperative control model data.

Description

BCM module cooperative control method based on voice instruction recognition under vehicle-mounted high-noise environment Technical Field The invention relates to the technical field of artificial intelligence, in particular to a BCM module cooperative control method based on voice instruction recognition in a vehicle-mounted high-noise environment. Background At present, due to the fact that noise components of a vehicle-mounted environment are complex and time-varying, when voice instruction recognition is carried out, the noise reduction algorithm has an inhibition effect on steady noise, but cannot completely adapt to nonstationary noise impact caused by sudden acceleration of an engine and sudden jolt of a road surface, when an input voice instruction is polluted by the sudden high-energy noise, serious distortion of voice characteristics can be caused, misjudgment is caused by a follow-up recognition model, and robustness of voice interaction in a high-noise environment cannot be guaranteed. Therefore, a BCM module cooperative control method based on voice command recognition in a vehicle-mounted high-noise environment is now proposed to solve the above-mentioned problems. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a BCM module cooperative control method based on voice instruction recognition in a vehicle-mounted high-noise environment, which solves the problems that the voice characteristics proposed in the background art are seriously distorted, so that the subsequent recognition model generates erroneous judgment, and the robustness of voice interaction in the high-noise environment cannot be ensured. In order to achieve the above purpose, the invention provides a BCM module cooperative control method based on voice instruction recognition in a vehicle-mounted high-noise environment, which comprises the following steps: s1, acquiring original voice instruction data and vehicle state parameter data in a vehicle-mounted high-noise environment; s2, preprocessing a voice signal in a noise environment based on the original voice instruction data to generate denoising voice instruction data; S3, voice instruction feature extraction processing is carried out according to the denoising voice instruction data, and voice instruction feature vector data are generated; S4, carrying out voice command recognition model matching processing based on the voice command feature vector data to generate preliminary voice command recognition result data; s5, performing BCM cooperative control strategy analysis by combining the vehicle state parameter data to generate BCM control instruction candidate set data; S6, performing instruction consistency verification processing based on the preliminary voice instruction recognition result data and the BCM control instruction candidate set data to generate optimized voice instruction recognition data; S7, performing BCM cooperative control instruction execution processing according to the optimized voice instruction identification data to generate BCM control execution feedback data; S8, performing performance evaluation and self-adaptive adjustment based on the BCM control execution feedback data to generate dynamic parameter update data; and S9, optimizing the voice recognition model and the BCM control strategy in real time by utilizing the dynamic parameter updating data, and generating self-adaptive cooperative control model data. Preferably, the step of collecting the original voice command data and the vehicle state parameter data in the vehicle-mounted high-noise environment in the step of S1 includes the following steps: s11, acquiring original voice instruction data in a high-noise environment in a vehicle through a vehicle-mounted microphone array, wherein the high-noise environment comprises engine noise, wind noise and road noise; S12, acquiring vehicle state parameter data including vehicle speed, engine speed, vehicle window state and air conditioner operation parameters through a vehicle CAN bus; S13, synchronizing the time stamp of the original voice instruction data and the vehicle state parameter data, and storing the time stamp to an on-board embedded database. Preferably, the step of preprocessing the voice signal in the noise environment in S2 includes the following steps: s21, performing background noise suppression processing on the original voice instruction data by adopting an adaptive filter to generate preliminary denoising voice data; S22, carrying out residual noise elimination on the primary denoising voice data based on spectral subtraction to generate secondary denoising voice data; S23, extracting effective voice segments through a voice activity detection algorithm, and generating denoising voice instruction data, wherein a voice activity detection threshold value is dynamically adjusted according to vehicle state parameters. Preferably, the voice instruction feature extraction processing in S3 incl