CN-122024760-A - Abnormality identification method, system and storage medium based on elevator intercom audio analysis

CN122024760ACN 122024760 ACN122024760 ACN 122024760ACN-122024760-A

Abstract

The invention relates to the technical field of voice analysis, in particular to an abnormality recognition method, a system and a storage medium based on elevator intercom audio analysis, which aim at the problems that the existing elevator abnormality recognition manual inspection efficiency is low, the cost of a single sensor is high, and the signal distortion and interference are difficult to solve in the traditional audio analysis, the method firstly obtains the original audio signal containing equipment frequency response distortion and environmental interference of an elevator intercom system, and obtaining a pure audio signal through signal quality compensation and interference noise suppression, extracting a Mel frequency cepstrum coefficient and discrete wavelet transformation characteristics from the pure signal, generating sensitive characteristics by combining the baseline characteristics under the health state of the elevator, comparing and diagnosing the fault type and position through a fault spectrum knowledge base, and updating an abnormal recognition model and the knowledge base based on operation and maintenance feedback. According to the invention, additional sensors are not required, the fault identification accuracy and instantaneity are improved, the model self-adaptive optimization is realized, the operation and maintenance cost is reduced, and the safe operation of the elevator is ensured.

Inventors

LUO XINGREN
FU SHIQIANG
YANG MINGHUI
CHENG WENBIN
DU CHENXI

Assignees

上海昶屹机电科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251229

Claims (10)

1. The abnormal identification method based on the elevator intercom audio analysis is characterized by comprising the following steps of: Acquiring an original audio signal of an elevator intercom system, wherein the original audio signal comprises noise and quality damage information caused by equipment frequency response characteristics and environmental interference; performing signal preprocessing and enhancement on the original audio signal, and obtaining a pure audio signal through signal quality compensation and interference noise suppression; Extracting sensitive features from the pure audio signals, wherein the sensitive features are used for representing abnormal sound characteristics in the running state of the elevator; Establishing a fault spectrum knowledge base, and performing fault diagnosis by comparing abnormal sound characteristics with the fault spectrum knowledge base so as to identify the abnormal type and fault position of the elevator; And updating the abnormal recognition model and the fault spectrum knowledge base according to the fault diagnosis result so as to realize continuous optimization and self-adaptive learning of the model.
2. The abnormality identification method based on elevator intercom audio analysis according to claim 1, wherein the step of performing signal preprocessing and enhancement on the original audio signal and obtaining a clean audio signal through signal quality compensation and interference noise suppression processing comprises: Performing signal quality compensation on an original audio signal, and performing reverse compensation on input audio by applying a pre-measured intercom system gain coefficient so as to correct signal distortion caused by equipment frequency response characteristics and preliminarily restore original characteristics of sound; Performing interference noise suppression on the audio signal subjected to signal quality compensation, filtering specific periodic noise by adopting an adaptive filtering technology, estimating the frequency spectrum of background noise through spectral subtraction, and subtracting the noise spectrum from the noisy signal; And verifying the quality of the preprocessed audio signal, evaluating the purity of the signal through the signal-to-noise ratio and the distortion index, and obtaining a pure audio signal.
3. The elevator intercom audio analysis based anomaly identification method of claim 1 wherein extracting a sensitive feature from the clean audio signal, the sensitive feature being used to characterize an abnormal sound characteristic in an elevator operating condition comprises: Performing time-frequency analysis on the pure audio signal, and extracting characteristics based on a human ear auditory model by adopting a mel frequency cepstrum coefficient; Performing multi-resolution analysis on the audio signal by using discrete wavelet transformation, capturing abnormal sound signals of non-stationary and burst by simultaneously analyzing the signals in a time domain and a frequency domain, and extracting wavelet coefficients as characteristics; Under the health state of the elevator, acquiring an audio signal in an idle period, and extracting a standard frequency spectrum as a baseline characteristic for reference standard of subsequent abnormality detection; comparing the extracted characteristic of the current audio signal with the baseline characteristic, calculating a characteristic difference value, highlighting an abnormal part through normalization processing, and generating a characteristic vector for fault diagnosis; And performing dimension reduction and selection on the extracted features, removing redundant information, and reserving a feature subset which is most effective in abnormal sound identification.
4. The abnormality identification method based on elevator intercom audio analysis according to claim 1, wherein the step of establishing a fault spectrum knowledge base, performing fault diagnosis by comparing abnormal sound characteristics with the fault spectrum knowledge base to identify an abnormality type and a fault location of an elevator comprises: Establishing a fault frequency spectrum knowledge base, collecting audio samples of various known fault types, including traction machine bearing abrasion, door vane displacement and guide rail oil shortage, marking fault types and position information of the audio samples, and extracting corresponding frequency spectrum characteristics; Using a deep learning model, performing similarity matching on the currently extracted sensitive features and features in a fault frequency spectrum knowledge base, and calculating Euclidean distance or cosine similarity between the features as a matching index; outputting probability judgment of fault types and possible positions according to the matching result, and determining reliability of the diagnosis result by setting a confidence threshold; Aiming at the generalization problem of different elevator models, a migration learning technology is adopted, a model is pre-trained on a large-scale general sound data set, and then small amount of data of a specific elevator is used for fine adjustment so as to quickly adapt to a new scene; And the diagnosis results, including fault positioning accuracy and false alarm rate, are evaluated through multiple dimensions, and model parameters are optimized to improve the diagnosis precision.
5. The abnormality recognition method based on elevator intercom audio analysis according to claim 1, wherein the step of updating the abnormality recognition model and the failure spectrum knowledge base according to the failure diagnosis result to realize continuous optimization and adaptive learning of the model comprises: Receiving on-site overhaul feedback of operation and maintenance personnel, confirming whether a diagnosis result is correct or not, and marking feedback data as a verified sample; adding the verified samples into a fault spectrum knowledge base, and expanding the sample diversity and coverage of the knowledge base; retraining the fault diagnosis model by using the expanded knowledge base, and updating model parameters by an incremental learning technology; and periodically executing a model updating flow, and dynamically adjusting a model structure according to new data to realize continuous self-adaptive learning.
6. The abnormality recognition method based on elevator intercom audio analysis according to claim 5, wherein the step of updating the abnormality recognition model and the failure spectrum knowledge base according to the failure diagnosis result to realize continuous optimization and adaptive learning of the model further comprises: converting the fault diagnosis result into a readable format, including fault type description, position information and confidence score, and displaying the fault diagnosis result to operation and maintenance personnel through a visual interface; generating maintenance suggestions according to the diagnosis results, wherein the maintenance suggestions comprise preferentially processing high-confidence abnormality and periodically checking low-frequency faults and are integrated into an elevator maintenance management system; monitoring the running state of the elevator in real time, triggering an automatic alarm mechanism when serious abnormality is identified, and outputting alarm information; And counting historical diagnosis data, analyzing abnormal occurrence rules and trends, and providing data support for preventive maintenance.
7. Abnormal recognition system based on elevator audio analysis that talkbacks, its characterized in that includes: the system comprises an acquisition module, a control module and a control module, wherein the acquisition module is used for acquiring an original audio signal of an elevator intercom system, and the original audio signal contains noise and quality damage information caused by equipment frequency response characteristics and environmental interference; The preprocessing module is used for carrying out signal preprocessing and enhancement on the original audio signal, and obtaining a pure audio signal through signal quality compensation and interference noise suppression processing; The extraction module is used for extracting sensitive features from the pure audio signals, and the sensitive features are used for representing abnormal sound characteristics in the running state of the elevator; the building module is used for building a fault frequency spectrum knowledge base, and carrying out fault diagnosis by comparing abnormal sound characteristics with the fault frequency spectrum knowledge base so as to identify the abnormal type and the fault position of the elevator; and the optimization module is used for updating the abnormal recognition model and the fault spectrum knowledge base according to the fault diagnosis result so as to realize continuous optimization and self-adaptive learning of the model.
8. The elevator intercom audio analysis based anomaly identification system of claim 7 wherein the preprocessing module comprises: the correction unit is used for carrying out signal quality compensation on the original audio signal, carrying out reverse compensation on the input audio by applying a pre-measured gain coefficient of the intercom system so as to correct signal distortion caused by the frequency response characteristic of equipment and preliminarily restore the original characteristics of sound; The noise suppression unit is used for performing interference noise suppression on the audio signal subjected to signal quality compensation, filtering specific periodic noise by adopting an adaptive filtering technology, estimating the frequency spectrum of background noise through spectral subtraction, and subtracting the noise spectrum from the noisy signal; the acquisition unit is used for verifying the quality of the preprocessed audio signal, evaluating the purity of the signal through the signal-to-noise ratio and the distortion index, and acquiring a pure audio signal.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 6 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.

Description

Abnormality identification method, system and storage medium based on elevator intercom audio analysis Technical Field The invention relates to the technical field of voice analysis, in particular to an abnormality identification method, an abnormality identification system and a storage medium based on elevator intercom audio analysis. Background As an indispensable vertical transportation means in modern buildings, the operation safety and reliability of elevators are directly related to personnel travel safety and building operation efficiency. With the increase of the service life of the elevator, core components such as a traction machine, a door machine, a guide rail and the like are easy to generate faults due to the problems of abrasion, oil shortage, displacement and the like, if the problems of abrasion, oil shortage, displacement and the like are not recognized and processed in time, the elevator is stopped to influence travel if the problems are light, and safety accidents such as passenger trapping of an elevator car, component damage and the like are caused if the problems are heavy. Therefore, the early abnormal recognition and accurate positioning of the elevator faults are realized, and the method is one of the core requirements in the elevator operation and maintenance field. The existing elevator abnormality recognition technology has defects in real-time, accuracy, universality and cost control, for example, the existing technology cannot effectively solve the core problems of difficult recognition of early faults, difficult distinction of fault types and low operation and maintenance efficiency in elevator operation and maintenance. Therefore, a technical scheme based on the existing elevator equipment, which is unnecessary to additionally install a sensor, can effectively process the distortion and interference of the audio signal, and realize accurate fault identification and continuous model optimization is needed, so that the reliability and operation and maintenance efficiency of elevator abnormal identification are improved, and the safe operation of the elevator is ensured. Disclosure of Invention The invention aims to provide an abnormality identification method based on elevator intercom audio analysis, which comprises the following steps: Acquiring an original audio signal of an elevator intercom system, wherein the original audio signal comprises noise and quality damage information caused by equipment frequency response characteristics and environmental interference; performing signal preprocessing and enhancement on the original audio signal, and obtaining a pure audio signal through signal quality compensation and interference noise suppression; Extracting sensitive features from the pure audio signals, wherein the sensitive features are used for representing abnormal sound characteristics in the running state of the elevator; Establishing a fault spectrum knowledge base, and performing fault diagnosis by comparing abnormal sound characteristics with the fault spectrum knowledge base so as to identify the abnormal type and fault position of the elevator; And updating the abnormal recognition model and the fault spectrum knowledge base according to the fault diagnosis result so as to realize continuous optimization and self-adaptive learning of the model. Further, the step of performing signal preprocessing and enhancement on the original audio signal and performing signal quality compensation and interference noise suppression processing to obtain a clean audio signal includes: Performing signal quality compensation on an original audio signal, and performing reverse compensation on input audio by applying a pre-measured intercom system gain coefficient so as to correct signal distortion caused by equipment frequency response characteristics and preliminarily restore original characteristics of sound; Performing interference noise suppression on the audio signal subjected to signal quality compensation, filtering specific periodic noise by adopting an adaptive filtering technology, estimating the frequency spectrum of background noise through spectral subtraction, and subtracting the noise spectrum from the noisy signal; And verifying the quality of the preprocessed audio signal, evaluating the purity of the signal through the signal-to-noise ratio and the distortion index, and obtaining a pure audio signal. Further, extracting a sensitive feature from the clean audio signal, the sensitive feature being used for characterizing abnormal sound characteristics in an elevator operating state, the step comprising: Performing time-frequency analysis on the pure audio signal, and extracting characteristics based on a human ear auditory model by adopting a mel frequency cepstrum coefficient; Performing multi-resolution analysis on the audio signal by using discrete wavelet transformation, capturing abnormal sound signals of non-stationary and burst by simultaneously analyzing the