阅读说明：本技术 基于压缩神经网络的设备异常工况声纹分析算法 (Equipment abnormal condition voiceprint analysis algorithm based on compressed neural network ) 是由 张利君 孔繁清 彭贵全 于 2021-06-23 设计创作，主要内容包括：一种基于压缩神经网络的设备异常工况声纹分析算法,包括设备异常工况声纹分析系统、设备运行情况的视频数据和变量数据；设备异常工况声纹分析系统包括：正常设备声纹数据库,异常设备声纹数据,设备音频采集单元实时采集设备运行音频,并通过音频分析工具将所采集音频进行音频图谱化；声纹匹配单元对设备运行声纹进行声纹匹配,实时监测设备运行状态：当设备运行声纹与正常设备声纹数据库内对应声纹一致时,判定该设备无异常：将该设备运行声纹实时更新至正常设备声纹数据库内；当设备运行声纹与正常设备声纹数据库内对应声纹不一致时,判定该设备异常并发出预警信号。(A voiceprint analysis algorithm for abnormal working conditions of equipment based on a compressed neural network comprises a voiceprint analysis system for the abnormal working conditions of the equipment, video data and variable data of the running conditions of the equipment; the equipment abnormal condition voiceprint analysis system comprises: the voice print data base of the normal equipment and the voice print data of the abnormal equipment, the equipment audio acquisition unit acquires the running audio of the equipment in real time and performs audio mapping on the acquired audio through an audio analysis tool; the voiceprint matching unit performs voiceprint matching on the voiceprint of the running equipment, and monitors the running state of the equipment in real time: when the operating voiceprint of the equipment is consistent with the corresponding voiceprint in the normal equipment voiceprint database, judging that the equipment is abnormal: updating the voice print of the equipment operation to a normal equipment voice print database in real time; and when the operating voiceprint of the equipment is inconsistent with the corresponding voiceprint in the normal equipment voiceprint database, judging that the equipment is abnormal and sending out an early warning signal.)

1. A voiceprint analysis algorithm for abnormal working conditions of equipment based on a compressed neural network is characterized in that: the method comprises an equipment abnormal working condition voiceprint analysis system, video data and variable data of equipment running conditions;

the equipment abnormal condition voiceprint analysis system comprises:

the normal equipment voiceprint database is internally stored with voiceprint atlas data of equipment in normal operation;

the abnormal equipment voiceprint database stores voiceprint atlas data of the abnormal operation of the equipment;

the equipment audio acquisition unit acquires equipment running audio in real time by using an online sound collection tool and performs audio mapping on the acquired audio through an audio analysis tool;

the voiceprint matching unit is used for carrying out voiceprint matching on the voiceprint of the equipment operation based on the normal equipment voiceprint database, and monitoring the equipment operation state in real time:

when the operating voiceprint of the equipment is consistent with the corresponding voiceprint in the normal equipment voiceprint database, judging that the equipment is abnormal: updating the voice print of the equipment operation to a normal equipment voice print database in real time; when the operating voiceprint of the equipment is inconsistent with the corresponding voiceprint in the normal equipment voiceprint database, judging that the equipment is abnormal and sending out an early warning signal;

the video data are input into a network model architecture composed of a compressed neural network and a fully connected neural network, the variable data are directly input into the fully connected neural network to obtain the identification results of the video data and the variable data, and after deep learning is carried out on the basis of big data composed of the video data and the variable data, the parameter setting of the network model architecture is completed to obtain a target network model architecture; the method comprises the steps of obtaining video data to be detected and variable data to be detected, inputting the video data to be detected and the variable data to be detected into a target network model architecture, detecting whether equipment operation conditions in the data to be detected are abnormal or not, classifying the equipment operation conditions if the equipment operation conditions are abnormal, and sending alarm indication information.

2. The compressed neural network-based equipment abnormal condition voiceprint analysis algorithm according to claim 1, wherein the voiceprint analysis algorithm comprises the following steps:

the method comprises the following steps that video data are input into a network model architecture formed by a compression neural network and a full-connection neural network, variable data are directly input into the full-connection neural network, and identification results of the video data and the variable data are obtained, and the method specifically comprises the following steps: inputting the video data into a compressed neural network, processing the video data through a convolutional layer and a pooling layer to obtain compressed data with reduced characteristic dimension, and inputting the compressed data into a full-connection layer of the full-connection neural network; obtaining variable data corresponding to the video data, adding the variable data into a full connection layer of a compressed neural network, and splicing the variable data with the compressed data to obtain target data; and carrying out operation processing on the target data through a hidden layer of the fully-connected neural network to obtain the identification results of the video data and the variable data.

Technical Field

The invention belongs to the technical field of artificial intelligence detection of data, and particularly relates to a voiceprint analysis algorithm for abnormal working conditions of equipment based on a compressed neural network.

Background

With the rapid development of artificial intelligence technology and the increase of computer hardware computing power, image processing systems with deep neural network technology as a core have come into existence, and deep learning models represented by deep neural networks have excellent performance in tasks such as target object detection and tracking, action recognition and the like, and are widely applied. The algorithm based on the deep neural network is increasingly applied to the field of intelligent detection, but most of the existing algorithms detect the appearance of parts at present, the processing object of the algorithm is an image, and the used network structure is a convolutional neural network. For products that are not visually identifiable as defective, the image-based approach described above cannot be used. In an actual production process, a worker can perform defect detection by recognizing abnormal sounds when an engine is operated, but in the field of computers, a technology for detecting defects of products by audio is still blank.

Most of the existing abnormal sound detection algorithms recognize the characteristics of abnormal sounds manually and summarize a set of algorithm flows to judge unknown sounds. The algorithm cannot automatically learn the characteristics of abnormal sounds, so that the application range is small, and the algorithm cannot be repeatedly used for different types of sound detection; some algorithms adopt a convolutional neural network to process audio, so that although the characteristics of abnormal sounds can be learned and detected by self, the convolutional neural network requires that the input has the same size, and the audio cannot meet the requirement generally; if the audio is pre-processed by cropping or the like, part of the information may be lost.

Based on the method, the voiceprint analysis algorithm of the abnormal working condition of the equipment based on the compressed neural network is designed to solve the problem that certain faults in the assembled equipment cannot be identified through observation.

Disclosure of Invention

The technical problem to be solved by the invention is to provide a voiceprint analysis algorithm for equipment under abnormal working conditions based on a compressed neural network, and combine an extraction algorithm for key fragments of the voiceprint of the equipment and an abnormal sound detection algorithm for deep learning to solve the problems in the background technology.

In order to solve the technical problem, the invention is realized by the following modes:

a voiceprint analysis algorithm for abnormal working conditions of equipment based on a compressed neural network comprises a voiceprint analysis system for the abnormal working conditions of the equipment, video data and variable data of the running conditions of the equipment;

the equipment abnormal condition voiceprint analysis system comprises:

the normal equipment voiceprint database is internally stored with voiceprint atlas data of equipment in normal operation;

the abnormal equipment voiceprint database stores voiceprint atlas data of the abnormal operation of the equipment;

the equipment audio acquisition unit acquires equipment running audio in real time by using an online sound collection tool and performs audio mapping on the acquired audio through an audio analysis tool;

the voiceprint matching unit is used for carrying out voiceprint matching on the voiceprint of the equipment operation based on the normal equipment voiceprint database, and monitoring the equipment operation state in real time:

when the operating voiceprint of the equipment is consistent with the corresponding voiceprint in the normal equipment voiceprint database, judging that the equipment is abnormal: updating the voice print of the equipment operation to a normal equipment voice print database in real time; when the operating voiceprint of the equipment is inconsistent with the corresponding voiceprint in the normal equipment voiceprint database, judging that the equipment is abnormal and sending out an early warning signal;

the video data are input into a network model architecture composed of a compressed neural network and a fully connected neural network, the variable data are directly input into the fully connected neural network to obtain the identification results of the video data and the variable data, and after deep learning is carried out on the basis of big data composed of the video data and the variable data, the parameter setting of the network model architecture is completed to obtain a target network model architecture; the method comprises the steps of obtaining video data to be detected and variable data to be detected, inputting the video data to be detected and the variable data to be detected into a target network model architecture, detecting whether equipment operation conditions in the data to be detected are abnormal or not, classifying the equipment operation conditions if the equipment operation conditions are abnormal, and sending alarm indication information.

Compared with the prior art, the invention has the following beneficial effects:

the abnormal working condition detection algorithm can simultaneously process the acquired video data and the input variable data based on the constructed network model architecture, improves the accuracy of off-line learning and the efficiency of on-line diagnosis, and further improves the use experience of users. Meanwhile, a compressed neural network abnormal voiceprint analysis method based on deep learning is adopted, the characteristics of normal and various abnormal sounds can be automatically learned through a large number of samples of known labels, the characteristics of the sounds do not need to be analyzed manually, and the compressed neural network abnormal voiceprint analysis method is conveniently used for processing other similar problems without being modified in a large amount; adopting a recurrent neural network model capable of processing variable-length sequence input: a cyclic neural network which can accept variable-length sequences as input is used as an input layer, and a deep neural network model is constructed on the basis of the cyclic neural network, so that the problem that all the input is required to have the same scale and the length of the audio is not fixed in the traditional method for processing the audio by using a compressed neural network is successfully solved.

Drawings

Fig. 1 is a schematic view of a structural module of the present invention.

Detailed Description

The following detailed description of embodiments of the invention is provided in connection with the accompanying drawings and the examples.

As shown in fig. 1, an abnormal working condition voiceprint analysis algorithm of a device based on a compressed neural network includes a system for analyzing the voiceprint of the abnormal working condition of the device, video data of the running condition of the device and variable data;

the equipment abnormal condition voiceprint analysis system comprises:

the normal equipment voiceprint database is internally stored with voiceprint atlas data of equipment in normal operation;

the abnormal equipment voiceprint database stores voiceprint atlas data of the abnormal operation of the equipment;

the equipment audio acquisition unit acquires equipment running audio in real time by using an online sound collection tool and performs audio mapping on the acquired audio through an audio analysis tool;

the voiceprint matching unit is used for carrying out voiceprint matching on the voiceprint of the equipment operation based on the normal equipment voiceprint database, and monitoring the equipment operation state in real time:

when the operating voiceprint of the equipment is consistent with the corresponding voiceprint in the normal equipment voiceprint database, judging that the equipment is abnormal: updating the voice print of the equipment operation to a normal equipment voice print database in real time; when the operating voiceprint of the equipment is inconsistent with the corresponding voiceprint in the normal equipment voiceprint database, judging that the equipment is abnormal and sending out an early warning signal;

the video data are input into a network model architecture composed of a compressed neural network and a fully connected neural network, the variable data are directly input into the fully connected neural network to obtain the identification results of the video data and the variable data, and after deep learning is carried out on the basis of big data composed of the video data and the variable data, the parameter setting of the network model architecture is completed to obtain a target network model architecture; the method comprises the steps of obtaining video data to be detected and variable data to be detected, inputting the video data to be detected and the variable data to be detected into a target network model architecture, detecting whether equipment operation conditions in the data to be detected are abnormal or not, classifying the equipment operation conditions if the equipment operation conditions are abnormal, and sending alarm indication information.

Further, the inputting of the video data into a network model architecture composed of a compressed neural network and a fully-connected neural network, and the directly inputting of the variable data into the fully-connected neural network to obtain the recognition results of the video data and the variable data specifically include: inputting the video data into a compressed neural network, processing the video data through a convolutional layer and a pooling layer to obtain compressed data with reduced characteristic dimension, and inputting the compressed data into a full-connection layer of the full-connection neural network; obtaining variable data corresponding to the video data, adding the variable data into a full connection layer of a compressed neural network, and splicing the variable data with the compressed data to obtain target data; and carrying out operation processing on the target data through a hidden layer of the fully-connected neural network to obtain the identification results of the video data and the variable data.

The foregoing is illustrative of embodiments of the present invention and it will be further appreciated by those skilled in the art that various modifications may be made without departing from the principles of the invention and that such modifications are intended to be included within the scope of the appended claims.