WO2021129236A1 - 一种采用一维卷积神经网络的声学垃圾分类方法 - Google Patents

一种采用一维卷积神经网络的声学垃圾分类方法 Download PDF

Info

Publication number
WO2021129236A1
WO2021129236A1 PCT/CN2020/129658 CN2020129658W WO2021129236A1 WO 2021129236 A1 WO2021129236 A1 WO 2021129236A1 CN 2020129658 W CN2020129658 W CN 2020129658W WO 2021129236 A1 WO2021129236 A1 WO 2021129236A1
Authority
WO
WIPO (PCT)
Prior art keywords
garbage
neural network
convolutional neural
sound
layer
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2020/129658
Other languages
English (en)
French (fr)
Inventor
邹俊
陆刚
王渊彬
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University
Original Assignee
Zhejiang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University filed Critical Zhejiang University
Priority to US17/640,340 priority Critical patent/US11835489B2/en
Publication of WO2021129236A1 publication Critical patent/WO2021129236A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/14—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object using acoustic emission techniques
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/04—Analysing solids
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/36—Detecting the response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/42—Detecting the response signal, e.g. electronic circuits specially adapted therefor by frequency filtering or by tuning to resonant frequency
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4409—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison
    • G01N29/4436—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison with a reference signal
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4454—Signal recognition, e.g. specific values or portions, signal events, signatures
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4481—Neural networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; ELECTRIC HEARING AIDS; PUBLIC ADDRESS SYSTEMS
    • H04R1/00—Details of transducers, loudspeakers or microphones
    • H04R1/08—Mouthpieces; Microphones; Attachments therefor
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2291/00—Indexing codes associated with group G01N29/00
    • G01N2291/02—Indexing codes associated with the analysed material
    • G01N2291/023—Solids
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Definitions

  • the invention relates to a garbage classification method, in particular, to an acoustic garbage classification method using a one-dimensional convolutional neural network.
  • Municipal solid waste management includes waste generation, source separation, storage, collection, transportation and treatment. Separation of the source of waste is a crucial step, because the quality of the classification will directly lead to how the waste is treated in the future. At present, most automatic classification technologies are mainly used for batch and large-scale classification processing in waste treatment plants. Effective waste separation technology is still lacking.
  • One method is to manually extract features of sound data and use a shallow classifier for classification. This method is cumbersome and has low classification accuracy.
  • Another method is to convert one-dimensional sound data into two-dimensional images and use a two-dimensional convolutional neural network for classification. This method is computationally complex, resulting in poor real-time performance. Therefore, the acoustic garbage classification method based on one-dimensional convolutional neural network will classify garbage quickly, effectively and in real time, improve the garbage collection rate, and get rid of the predicament of garbage siege.
  • the purpose of the present invention is to provide an acoustic garbage classification method using a one-dimensional convolutional neural network, which uses sound to recognize and classify garbage, and mainly solves the problem of difficulty in man-made garbage classification in the process of garbage source separation, thereby achieving high accuracy Source classification of garbage.
  • (C) Collect sound signals of different types of garbage and preprocess it to build a sound database for garbage classification, establish and train a one-dimensional convolutional neural network model;
  • the method of the present invention for deep learning processing based on sound to realize garbage classification is helpful for accurate garbage classification, thereby improving the accuracy of garbage classification and recycling, and has high practical and popularization value.
  • the sound generated by the garbage falling is produced by hitting the garbage freely on a board, and the generated impact sound is recorded by a single-channel microphone; the sampling frequency is 44100Hz; the above process is repeated for each type of garbage. Obtain multiple sound signals at a time.
  • each sound signal only intercepts the effective duration of 120ms, and the sound signal of 120ms duration starts from the peak of each sound signal and ends 120ms backward, so that this sound signal covers During the entire impact process, redundant information in the sound signal can be removed.
  • the one-dimensional convolutional neural network model mainly includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer; when the network model is trained, the input layer is the preprocessing of the marked garbage type After the sound signal; the convolutional layer performs convolution operations on the output data of the previous layer, extracts features, uses the ReLU activation function to enhance the nonlinear performance of the convolutional neural network, and then passes through the maximum pooling layer to perform feature dimensionality reduction and reduction The number of network parameters and over-fitting are finally classified through the fully connected layer and the output layer and output the classification results.
  • the selection of different hyperparameters will affect the classification performance of the network; the present invention mainly considers three hyperparameters: network depth, convolution kernel size and learning Rate; considering the training time, computational complexity and computer configuration issues, only five empirical values are selected for each parameter; the orthogonal experiment method is used to obtain the optimal combination of three parameter values, which makes the classification of the one-dimensional convolutional neural network model Optimal performance.
  • the present invention creatively uses sound data to classify garbage.
  • Sound is highly correlated with material properties, and sound signals can reveal the inherent characteristics of garbage, such as elasticity and internal friction. Elasticity affects the frequency at which sound is produced. Internal friction determines how the generated sound decays over time and provides acoustic characteristics of different shapes for waste sorting. At the same time, the use of sound for garbage classification greatly reduces the data set for training the model.
  • the orthogonal experiment method is used to study the influence of different hyperparameters on the classification performance of the one-dimensional convolutional neural network, and finally determine the optimal hyperparameter combination, so that the network model Obtain the best classification performance.
  • the present invention uses a one-dimensional convolutional neural network to classify garbage.
  • the one-dimensional convolutional neural network is suitable for processing one-dimensional signals and can directly extract the optimal features from the original data, avoiding the cumbersomeness of manually extracting features.
  • the sub-optimality and the computational complexity of the two-dimensional convolutional neural network classification maximize the classification accuracy.
  • one-dimensional convolutional neural networks are suitable for real-time applications on mobile, low-power, low-memory devices.
  • Figure 1 is a flow chart of the method of the present invention
  • Figure 2 is a model diagram of a convolutional neural network for garbage classification based on sound signals.
  • Fig. 3 is a schematic diagram of the structure of the one-dimensional convolutional neural network of the present invention.
  • the specific implementation of the one-dimensional convolutional neural network model chooses TensorFlow 1.9 as the deep learning framework, all algorithms are written in python 3.6, and all programs are in a laptop with GTX1050Ti graphics card, 8th generation Intel Core i5 processor and 512GB SSD Executed on.
  • the learning rate is 1e-5, 5e-5, 1e-4, 5e-4, 1e-3, and 25 experiments are carried out according to the orthogonal design table of six factors and five levels, and finally get The optimal parameter combination is: the network depth is 5, the convolution kernel size is 9, and the learning rate is 5e-4.
  • the one-dimensional convolutional neural network model of the embodiment mainly includes an input layer, five convolutional layers, five pooling layers, a fully connected layer, and an output layer.
  • the one-dimensional sound signal is input to the input layer of the one-dimensional convolutional neural network.
  • a convolution operation is performed between the input signal and the corresponding convolution kernel to generate an input feature map.
  • a pooling layer is usually used, which not only reduces the computational cost by reducing the dimension of the features extracted from the convolutional layer, but also provides basic translation invariance to the features.
  • the specific design of the one-dimensional convolutional neural network model of the embodiment is as follows: define convolutional layers 1-5, the number of convolution kernels are 8, 16, 24, 32 and 40, the kernel size is 9 ⁇ 9, strides Is 1, the boundary padding is SAME, and the activation function is ReLU function; defines the maximum pooling layer 1-5, the number of convolution kernels are 8, 16, 24, 32 and 40 respectively, the kernel size is 9 ⁇ 9, and the stride strides is 2; finally, the fully connected layer and the output layer are defined, and the softmax function is used for classification, and the label 0-n of the category is output.
  • the training process is mainly divided into two stages: forward propagation and back propagation.
  • Forward propagation is to build the calculation process of the model and establish the mapping relationship between input and output;
  • back propagation is to use gradient descent method to train model parameters to minimize the loss function.
  • After obtaining the final model use the test set in the sound database to evaluate the classification effect of the model.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Immunology (AREA)
  • Health & Medical Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Pathology (AREA)
  • Signal Processing (AREA)
  • Acoustics & Sound (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Image Analysis (AREA)
  • Circuit For Audible Band Transducer (AREA)

Abstract

一种采用一维卷积神经网络的声学垃圾分类方法。采集垃圾掉落产生的声音信号；声音信号预处理；采集不同种类垃圾的声音信号并预处理搭建垃圾分类的声音数据库，建立并训练一维卷积神经网络模型；获取待分类垃圾的声，输入训练后的一维卷积神经网络进行垃圾分类并输出分类结果。垃圾分类方法有助于辅助人们进行准确地垃圾分类，从而提高垃圾分类回收的准确率，具有很高的实用和推广价值。

Description

一种采用一维卷积神经网络的声学垃圾分类方法 技术领域
本发明涉及垃圾分类方法，具体地讲，尤其是涉及了一种采用一维卷积神经网络的声学垃圾分类方法。
背景技术
由于人口、城市化和经济的快速发展，城市固体垃圾的数量在急剧增加，这给城市固体垃圾管理带来了巨大的挑战。城市固体垃圾管理包括垃圾的产生、源头分离、储存、收集、运输和处理等。垃圾源头分离是至关重要的一步，因为分类的质量将直接导致这些垃圾后续被如何处理。当前，大多数自动分类技术主要用于垃圾处理厂的批量、大规模分类处理。有效的垃圾源头分离技术仍然是缺乏的。
目前，垃圾的源头分离主要依靠人们的意识和具有的垃圾分类知识，这就需要花费大量的时间和金钱在教育和监督上。在许多发展中国家，让人们在短时间内形成垃圾分类的习惯是不现实的。因此，开发一种垃圾源头分离技术在人们的日常生活中帮助他们进行垃圾分类迫在眉睫。图像识别技术已经被广泛的运用在垃圾分类中。然而每一种垃圾可以有各种各样的表现形式，如不同的颜色、形状、变形、与其他物质的混合等。这给图像分类模型训练准备足够的、无偏倚的数据集带来了很大的挑战。物体的声音与它的图像不同，声音与材料属性高度相关，使用声音对垃圾进行分类是一个很好的选择。
如今，涉及声音分类的方法主要有两种。一种方法是人工提取声音数据特征，使用浅层分类器进行分类。这种方法过程繁琐，分类准确率低。另一种方法是将一维的声音数据转换为二维图像，使用二维卷积神经网络进行分类。这种方法计算复杂，导致实时性差。因此，基于一维卷积神经网络的声学垃圾分类方法将快速、有效、实时地对垃圾进行分类，提高垃圾回收率，摆脱垃圾围城困境。
发明内容
本发明的目的在于提供一种采用一维卷积神经网络的声学垃圾分类方法，利用声音进行垃圾的识别和分类，主要解决垃圾源头分离过程中人为垃圾分类困难的问题，从而实现高准确率的垃圾源头分类。
如图1所示，本发明采用的技术方案如下：
(A)采集垃圾掉落产生的声音信号；
(B)声音信号预处理；
(C)采集不同种类垃圾的声音信号并预处理搭建垃圾分类的声音数据库，建立并训练一维卷积神经网络模型；
(D)获取待分类垃圾的声，输入训练后的一维卷积神经网络进行垃圾分类并输出分类结果。
本发明基于声音进行深度学习处理进而实现垃圾分类的方法有助于准确地垃圾分类，从而提高垃圾分类回收的准确率，具有很高的实用和推广价值。
所述步骤(A)中，垃圾掉落产生的声音是将垃圾自由落体撞击到一块板上产生的，产生的撞击声被一个单通道麦克风记录；采样频率为44100Hz；每种垃圾重复上述过程多次获得多段声音信号。
所述步骤(B)中，预处理为：每段声音信号只截取120ms有效时长，120ms时长的声音信号以每段声音信号的峰值为起点向后120ms为终点，，这样使得这段声音信号覆盖整个撞击过程，并能去除声音信号中的冗余信息。
所述步骤(C)中，一维卷积神经网络模型主要包括输入层、卷积层、池化层、全连接层和输出层；网络模型训练时，输入层为已标注垃圾类型的预处理后的声音信号；卷积层对上一层的输出数据进行卷积运算、提取特征，采用ReLU激活函数以增强卷积神经网络的非线性性能，接着经过最大池化层进行特征降维、减少网络参数数量和过拟合，最后通过全连接层、输出层进行分类并输出分类结果。
所述步骤(C)中，一维卷积神经网络模型训练过程中，不同的超参数的选择将影响网络的分类性能；本发明主要考虑三个超参数：网络深度、卷积核大小和学习率；考虑到训练时间、计算复杂度和计算机配置问题，每个参数只选取五个经验值；采用正交实验法获得三个参数值的最优组合，使得一维卷积神经网络模型的分类性能最优。
本发明具有以下有益效果：
(1)本发明创造性地使用声音数据进行垃圾分类，声音与材料属性高度相关，声音信号可以揭示垃圾的固有特性，如弹性和内摩擦。弹性影响产生声音的频率。内摩擦决定了产生的声音如何随着时间衰减，并为垃圾分类提供了形状不同的声学特征。同时，使用声音进行垃圾分类大大减少了训练模型的数据集。
(2)本发明在一维卷积神经网络训练过程中，采用正交实验法研究不同超参数对一维卷积神经网络分类性能的影响，最终确定最优的超参数组合，从而使网络模型获得最优分类性能。
(3)本发明通过一维卷积神经网络进行垃圾分类，一维卷积神经网络适用于处理一维信号，能够直接从原始数据中提取到最优特征，避免了人工提取特征的繁琐性、次优性以及二维卷积神经网络分类的计算复杂性，最大限度地提高了分类准确度。同时，一维卷积神经网络适合于可移动、低功耗、低内存设备上的实时应用。
附图说明
图1为本发明的方法流程图；
图2为基于声音信号的垃圾分类卷积神经网络模型图。
图3为本发明的一维卷积神经网络的结构示意图。
具体实施方式
下面结合附图和实施例对本发明作进一步说明，本发明的实施方式包括但不限于下列实施例。
本发明的具体实施例及其实施情况如下：
具体实施的一维卷积神经网络模型选择TensorFlow 1.9作为深度学习框架，所有的算法用python 3.6编写，所有程序在一台带有GTX1050Ti显卡、第8代Intel Core i5处理器和512GB SSD的笔记本电脑上执行。
如图2所示，基于声音采集装置，准备各类垃圾材料，然后进行垃圾声音的采集与标注，对这些数据进行预处理，设计一种简单高效的网络结构来分析声音数据。卷积网络中包含了大量参数，采用正交实验法，搜索最优的参数组合；在正交实验过程中，网络深度分别取1、2、3、4、5，卷积核大小分别取3、5、7、9、11，学习率分别取1e-5、5e-5、1e-4、5e-4、1e-3，依据六因素五水平的正交设计表进行25次实验，最终获得最优的参数组合为：网络深度取5，卷积核大小取9，学习率取5e-4。
如图3所示，实施例的一维卷积神经网络模型主要包括输入层、五个卷积层、五个池化层、全连接层和输出层。一维声音信号被输入到一维卷积神经网络的输入层。在输入信号和对应的卷积核之间进行卷积运算，生成输入特征映射。然后通过激活函数传递输入特征映射，生成卷积层的输出特征映射。在卷积层之后，通常采用池化层，不仅可通过降低从卷积层提取的特征维数来减少计算成本，而且对特征提供了基本的平移不变性。
实施例的一维卷积神经网络模型的具体设计为：定义卷积层1-5，卷积核个数分别为8，16，24，32和40，核大小为9×9，步幅strides为1，边界填充padding为SAME，激活函数选用ReLU函数；定义最大池化层1-5，卷积核个数分别为8，16，24，32和40，核大小为9×9，步幅strides为2；最后定义全 连接层和输出层，并使用softmax函数进行分类，并输出所属类别的标签0-n。
训练及评估模型。训练过程主要分为前向传播和反向传播两个阶段。前向传播就是搭建了模型的计算过程，建立了输入与输出的映射关系；反向传播就是使用梯度下降法训练模型参数，实现损失函数最小。获得最终模型后，使用声音数据库中的测试集对模型分类效果进行评估。
(D)获取待分类垃圾的声音，输入训练后的一维卷积神经网络进行垃圾分类并输出分类结果。
以上所述仅为本发明的优先实施方式，本发明并不限定于上述实施方式，只要以基本相同手段实现本发明目的的技术方案都属于本发明的保护范围之内。

Claims (4)

  1. 一种采用一维卷积神经网络的声学垃圾分类方法，其特征在于：方法包括以下步骤：
    (A)采集垃圾掉落产生的声音信号；
    (B)声音信号预处理；
    (C)采集不同种类垃圾的声音信号并预处理搭建垃圾分类的声音数据库，建立并训练一维卷积神经网络模型；
    (D)获取待分类垃圾的声，输入训练后的一维卷积神经网络进行垃圾分类并输出分类结果。
  2. 根据权利要求书1所述的一种采用一维卷积神经网络的声学垃圾分类方法，其特征在于：所述步骤(A)中，垃圾掉落产生的声音是将垃圾自由落体撞击到一块板上产生的，产生的撞击声被一个单通道麦克风记录；采样频率为44100Hz；每种垃圾重复上述过程多次获得多段声音信号。
  3. 根据权利要求书1所述的一种采用一维卷积神经网络的声学垃圾分类方法，其特征在于：所述步骤(B)中，预处理为：每段声音信号只截取120ms有效时长，120ms时长的声音信号以每段声音信号的峰值为起点向后120ms为终点。
  4. 根据权利要求书1所述的一种采用一维卷积神经网络的声学垃圾分类方法，其特征在于：所述步骤(C)中，一维卷积神经网络模型主要包括输入层、卷积层、池化层、全连接层和输出层；网络模型训练时，输入层为已标注垃圾类型的预处理后的声音信号；卷积层对上一层的输出数据进行卷积运算、提取特征，采用ReLU激活函数以增强卷积神经网络的非线性性能，接着经过最大池化层进行特征降维、减少网络参数数量和过拟合，最后通过全连接层、输出层进行分类并输出分类结果。
PCT/CN2020/129658 2019-12-24 2020-11-18 一种采用一维卷积神经网络的声学垃圾分类方法 Ceased WO2021129236A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/640,340 US11835489B2 (en) 2019-12-24 2020-11-18 Acoustic garbage classification method using one-dimensional convolutional neural network (1D-CNN)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911347540.6 2019-12-24
CN201911347540.6A CN111160438A (zh) 2019-12-24 2019-12-24 一种采用一维卷积神经网络的声学垃圾分类方法

Publications (1)

Publication Number Publication Date
WO2021129236A1 true WO2021129236A1 (zh) 2021-07-01

Family

ID=70558143

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/129658 Ceased WO2021129236A1 (zh) 2019-12-24 2020-11-18 一种采用一维卷积神经网络的声学垃圾分类方法

Country Status (3)

Country Link
US (1) US11835489B2 (zh)
CN (1) CN111160438A (zh)
WO (1) WO2021129236A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116550635A (zh) * 2023-04-12 2023-08-08 深圳康佳电子科技有限公司 一种基于声纹识别的可回收垃圾分类方法及相关设备
CN120971585A (zh) * 2025-08-01 2025-11-18 中国电建集团西北勘测设计研究院有限公司 一种基于深度学习融合特征的pccp断丝监测识别分析方法

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111160438A (zh) * 2019-12-24 2020-05-15 浙江大学 一种采用一维卷积神经网络的声学垃圾分类方法
SMT202500393T1 (it) 2020-05-31 2025-11-10 Candam Tech S L Metodo per classificare rifiuti
CN113851115A (zh) * 2021-09-07 2021-12-28 中国海洋大学 一种基于一维卷积神经网络的复杂声音识别方法
CN114324580A (zh) * 2021-12-03 2022-04-12 西安交通大学 一种结构缺陷的智能敲击检测方法及系统
CN116975627B (zh) * 2023-06-20 2025-11-14 大连交通大学 一种声发射信号识别方法及网络模型训练方法
CN117465862A (zh) * 2023-12-08 2024-01-30 中通服科信信息技术有限公司 智能垃圾桶管理方法和装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1405706A1 (en) * 2002-10-02 2004-04-07 Nippon Steel Chemical Co., Ltd. Method of recycling waste plastic
KR20190071387A (ko) * 2017-12-14 2019-06-24 안성준 재활용 쓰레기 자동 분류 장치 및 방법
CN110427896A (zh) * 2019-08-07 2019-11-08 成都理工大学 一种基于卷积神经网络的垃圾分类智能系统
CN110937280A (zh) * 2019-12-24 2020-03-31 浙江大学 一种基于视听结合的智能垃圾分类回收方法和终端
CN111160438A (zh) * 2019-12-24 2020-05-15 浙江大学 一种采用一维卷积神经网络的声学垃圾分类方法

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101391693A (zh) * 2008-10-14 2009-03-25 清华大学 自动分类的垃圾桶
CN107416387B (zh) * 2017-07-19 2020-07-07 浙江联运环境工程股份有限公司 智能分类垃圾箱
CN108932480B (zh) * 2018-06-08 2022-03-15 电子科技大学 基于1d-cnn的分布式光纤传感信号特征学习与分类方法
CN109606991B (zh) * 2019-01-11 2022-02-08 郑州大学 智能垃圾桶及基于深度学习的垃圾分类方法
CN110070888A (zh) * 2019-05-07 2019-07-30 颐保医疗科技(上海)有限公司 一种基于卷积神经网络的帕金森语音识别方法
CN110598800A (zh) * 2019-09-23 2019-12-20 山东浪潮人工智能研究院有限公司 一种基于人工智能的垃圾分类识别方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1405706A1 (en) * 2002-10-02 2004-04-07 Nippon Steel Chemical Co., Ltd. Method of recycling waste plastic
KR20190071387A (ko) * 2017-12-14 2019-06-24 안성준 재활용 쓰레기 자동 분류 장치 및 방법
CN110427896A (zh) * 2019-08-07 2019-11-08 成都理工大学 一种基于卷积神经网络的垃圾分类智能系统
CN110937280A (zh) * 2019-12-24 2020-03-31 浙江大学 一种基于视听结合的智能垃圾分类回收方法和终端
CN111160438A (zh) * 2019-12-24 2020-05-15 浙江大学 一种采用一维卷积神经网络的声学垃圾分类方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
LIU QIANG, CHEN FU-LAN: "Research on Classification Methods for Marine Life Rubbish Based on Cepstrum and BP Network", JOURNAL OF NANTONG VOCATIONAL & TECHNICAL SHIPPING COLLEGE, vol. 15, no. 1, 1 March 2016 (2016-03-01), pages 34 - 38, XP055824738, DOI: 10.3969/j.issn.1671—9891.2016.01.010 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116550635A (zh) * 2023-04-12 2023-08-08 深圳康佳电子科技有限公司 一种基于声纹识别的可回收垃圾分类方法及相关设备
CN120971585A (zh) * 2025-08-01 2025-11-18 中国电建集团西北勘测设计研究院有限公司 一种基于深度学习融合特征的pccp断丝监测识别分析方法

Also Published As

Publication number Publication date
US11835489B2 (en) 2023-12-05
US20220349861A1 (en) 2022-11-03
CN111160438A (zh) 2020-05-15

Similar Documents

Publication Publication Date Title
US11835489B2 (en) Acoustic garbage classification method using one-dimensional convolutional neural network (1D-CNN)
CN107122375B (zh) 基于图像特征的图像主体的识别方法
CN108763216A (zh) 一种基于中文数据集的文本情感分析方法
CN102722713B (zh) 一种基于李群结构数据的手写体数字识别方法及系统
CN112733936A (zh) 一种基于图像识别的可回收垃圾分类方法
CN111860171A (zh) 一种大规模遥感图像中不规则形状目标的检测方法及系统
CN111695392A (zh) 基于级联的深层卷积神经网络的人脸识别方法及系统
CN104361342A (zh) 一种基于几何不变形状特征的在线植物物种识别方法
CN111368649B (zh) 一种运行于树莓派的情绪感知方法
Li et al. Dating ancient paintings of Mogao Grottoes using deeply learnt visual codes
CN113069117A (zh) 一种基于时间卷积神经网络的脑电情绪识别方法及系统
CN108596038A (zh) 一种结合形态学分割和神经网络的粪便中红细胞识别方法
CN118262746B (zh) 智能化噪声识别方法
CN110554429A (zh) 一种基于变邻域滑动窗口机器学习的地震断层识别方法
CN109815920A (zh) 基于卷积神经网络和对抗卷积神经网络的手势识别方法
CN109086794B (zh) 一种基于t-lda主题模型的驾驶行为模式识方法
Yingxin et al. A robust hand gesture recognition method via convolutional neural network
Kailash et al. Deep learning based detection of mobility aids using yolov5
CN116451114A (zh) 基于企业多源实体特征信息的物联网企业分类系统及方法
CN115115932A (zh) 一种基于垃圾目标检测的图像识别改进方法
CN109471941A (zh) 一种应对类别不平衡的罪名分类方法
CN112507863B (zh) 基于量子Grover算法的手写文字图片分类方法
CN110046576A (zh) 一种训练识别面部表情的方法和装置
CN116363732B (zh) 人脸情绪识别方法、装置、设备及存储介质
CN118429980A (zh) 基于深度学习的书法临帖效果评价方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20904894

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20904894

Country of ref document: EP

Kind code of ref document: A1