WO2020143512A1 - 一种婴儿哭声识别方法、装置及设备 - Google Patents

一种婴儿哭声识别方法、装置及设备 Download PDF

Info

Publication number
WO2020143512A1
WO2020143512A1 PCT/CN2019/130824 CN2019130824W WO2020143512A1 WO 2020143512 A1 WO2020143512 A1 WO 2020143512A1 CN 2019130824 W CN2019130824 W CN 2019130824W WO 2020143512 A1 WO2020143512 A1 WO 2020143512A1
Authority
WO
WIPO (PCT)
Prior art keywords
sequence
feature
feature sequence
recognition
voice data
Prior art date
Application number
PCT/CN2019/130824
Other languages
English (en)
French (fr)
Inventor
乔宇
王群
Original Assignee
深圳先进技术研究院
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 深圳先进技术研究院 filed Critical 深圳先进技术研究院
Publication of WO2020143512A1 publication Critical patent/WO2020143512A1/zh

Links

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/18—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being spectral information of each sub-band
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/21—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being power information
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/24—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being the cepstrum
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
    • G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
    • G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination

Definitions

  • the present application belongs to the field of voice recognition, and particularly relates to a method, device and equipment for baby cry recognition.
  • the current common identification scheme is based on a single indicator such as decibel, zero-crossing rate or energy to determine whether an alarm is required. When interference noise occurs in the environment, it is easy to make the false alarm rate higher.
  • the embodiments of the present application provide a method, device and equipment for recognizing baby crying, in order to solve the problem of the prior art method for recognizing baby crying, when the environment is interfered, it is easy to make the false alarm rate high .
  • a first aspect of the embodiments of the present application provides a method for identifying crying of a baby.
  • the method for identifying crying of a baby includes:
  • Identify the audio feature vector according to a preset recognition model and send the recognition result to the monitoring terminal.
  • the step of calculating two or more audio feature vectors included in the voice data segment includes:
  • Two or more feature sequences of the zero-crossing rate feature sequence, the energy feature sequence, the multi-order Mel frequency cepstrum coefficient feature sequence or the spectral centroid feature sequence are selected to generate the audio feature vector.
  • the selected zero-crossing rate characteristic sequence, energy characteristic sequence, multi-order Mel frequency cepstrum coefficient characteristic sequence or spectrum include:
  • the audio feature vector is determined according to the calculated average.
  • the step of identifying the audio feature vector according to a preset identification model and sending the identification result to the monitoring terminal includes:
  • the audio feature vector is sent to the cloud server, so that the cloud server sends an application reminder message to the monitoring terminal according to the recognition result.
  • the method further includes:
  • the audio feature vector is identified through a locally stored neural network model
  • the recognition result is a predetermined alarm result
  • the method before the step of calculating two or more audio feature vectors included in the voice data segment, the method further includes:
  • a second aspect of an embodiment of the present application provides a baby cry recognition device, the baby cry recognition device includes:
  • Voice data collection unit used to collect voice data and intercept voice data segments of a predetermined duration
  • An audio feature vector calculation unit configured to calculate two or more audio feature vectors included in the speech data segment
  • the recognition unit is configured to recognize the audio feature vector according to a preset recognition model, and send the recognition result to the monitoring terminal.
  • the audio feature vector calculation unit includes:
  • a feature sequence calculation subunit configured to calculate two or more of a zero-crossing rate feature sequence, an energy feature sequence, a multi-order Mel frequency cepstrum coefficient feature sequence or a spectral centroid feature sequence in the voice data segment;
  • a selection subunit is used to select two or more feature sequences of a zero-crossing rate feature sequence, an energy feature sequence, a multi-order Mel frequency cepstrum coefficient feature sequence, or a spectral centroid feature sequence to generate an audio feature vector.
  • a third aspect of an embodiment of the present application provides a baby cry recognition device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the The computer program implements the steps of the baby cry recognition method according to any one of the first aspect.
  • a fourth aspect of the embodiments of the present application provides a computer-readable storage medium that stores a computer program, and the computer program is executed by a processor to implement an infant as described in any one of the first aspects The steps of the cry recognition method.
  • the beneficial effects of the embodiments of the present application are: by collecting voice data, intercepting a voice data segment of a predetermined duration, calculating two or more audio feature vectors included in the voice data segment, according to The preset recognition model recognizes the audio feature vector and sends the recognition result to the monitoring terminal. Since the recognition result is recognized based on two or more audio feature vectors, the recognition result is more accurate and reliable, which is beneficial to improve the accuracy of baby cry recognition.
  • FIG. 1 is a schematic diagram of an implementation scenario of a method for identifying a crying baby according to an embodiment of the present application
  • FIG. 2 is a schematic flowchart of an implementation of a method for identifying a crying baby according to an embodiment of the present application
  • FIG. 3 is a schematic diagram of an implementation process of yet another baby cry recognition method provided by an embodiment of the present application.
  • FIG. 4 is a schematic diagram of a baby cry recognition device provided by an embodiment of the present application.
  • FIG. 5 is a schematic diagram of a baby cry recognition device provided by an embodiment of the present application.
  • FIG. 1 is a schematic diagram of an implementation scenario of a baby cry recognition method provided by an embodiment of the present application.
  • the implementation scenario includes an acquisition terminal, a monitoring terminal, and a cloud server.
  • the collection terminal may be a smart phone, a tablet computer, etc.
  • the application program can be installed on the collection terminal without additional hardware equipment, which is helpful to reduce the hardware cost of crying baby recognition.
  • the collection terminal is used to adopt the voice data of the baby, and segment the captured voice data according to a preset time length. For example, taking the current time as a starting point, intercepting the voice data collected within a predetermined period of time before the current time to obtain a voice data segment.
  • the predetermined duration may be 30 seconds or the like.
  • the voice data segment may be sent to a cloud server, and the cloud server calculates the collected voice data segment, and recognizes that the voice data segment includes Of two or more audio feature vectors to generate a baby cry recognition result.
  • the collection terminal has no network connection, it can also be recognized by the recognition model stored locally by the collection terminal.
  • the monitoring terminal is a device carried by the monitoring personnel, and may be a smart phone or the like. You can receive reminders through the installed applications, or by SMS or phone.
  • FIG. 2 is a schematic diagram of an implementation process of a method for recognizing crying baby provided by an embodiment of the present application, which is described in detail as follows:
  • step S201 collect voice data and intercept voice data segments of a predetermined duration
  • the baby cry recognition method described in this application may be implemented based on existing smart devices. Installing the application corresponding to the baby cry recognition method described in this application in a smart device equipped with a microphone can effectively analyze and process the collected voice data to obtain the recognition result of the baby cry.
  • the voice data can be segmented and intercepted according to a preset predetermined duration.
  • the voice data segment may be obtained by taking voice data of a predetermined length from the current time according to the current time being the end time of the voice data segment.
  • the interception time interval of the voice data segment may be set according to a predetermined duration of the voice data segment. For example, a certain proportion of the predetermined duration of the voice data segment can be taken. For example, when the predetermined duration is 30 seconds, the interception time of the voice data segment may be 5 seconds, etc., so that the voice data can be dynamically analyzed.
  • step S202 calculate two or more audio feature vectors included in the voice data segment
  • the audio feature vector included in the speech data segment may include two or more of a zero-crossing rate feature sequence, an energy feature sequence, a multi-order Mel frequency cepstrum coefficient feature sequence, or a spectral centroid feature sequence.
  • the characteristic sequence of multi-order Mel frequency cepstrum coefficients may be a characteristic sequence of 13-order Mel frequency cepstrum coefficients.
  • the energy may be a numerical value of the energy change.
  • the multi-order mel frequency cepstrum is a linear transformation of the log energy spectrum based on the non-linear mel scale of the sound frequency.
  • Mel frequency cepstrum coefficient (Mel-Frequency Cepstral Coefficients, MFCCs) are the coefficients that make up the Mel frequency cepstrum. It is derived from the cepstrum of audio clips.
  • a preferred embodiment is to select a multi-dimensional feature sequence including a zero-crossing rate feature sequence, an energy feature sequence, a multi-order Mel frequency cepstrum coefficient feature sequence, and a spectral centroid feature sequence.
  • a 16-dimensional characteristic sequence can be selected, so as to facilitate obtaining a more accurate recognition result.
  • two or more feature sequences of a zero-crossing rate feature sequence, an energy feature sequence, a multi-order Mel frequency cepstral coefficient feature sequence, or a spectral centroid feature sequence can be selected to directly obtain audio Feature vector.
  • the audio feature vector may also be calculated by selecting two or more feature sequences from a zero-crossing rate feature sequence, an energy feature sequence, a multi-order Mel frequency cepstrum coefficient feature sequence, or a spectral centroid feature sequence
  • the mean value of the selected feature sequence is determined according to the calculated mean value.
  • one or more steps of emphasizing, framing, and windowing the voice data segment may also be included. among them:
  • this coefficient can be The frequency is positively correlated, so that the amplitude of the high frequency will be increased.
  • the voice signal is not stable macroscopically, it has short-term stability microscopically (the voice signal can be considered to be approximately unchanged within 10-30 ms).
  • the voice signal can be divided into short segments , That is, processing is divided into frames, and each short segment is called a frame.
  • the speech data segment can also be windowed, that is, the speech data is multiplied by a window function to make the global more continuous and avoid the Gibbs effect.
  • the original non-periodic speech signal exhibits some features of the periodic function.
  • step S203 the audio feature vector is recognized according to a preset recognition model, and the recognition result is sent to the monitoring terminal.
  • the recognition model may be a neural network model.
  • a large number of baby crying samples and noise samples can be collected, audio feature vectors of crying samples and noise samples are calculated, the neural network model is trained, and the audio feature vectors are identified and processed according to the trained neural network model .
  • crying data of a monitored baby may be collected, and the recognition model may be trained to enable a more reliable recognition result.
  • the recognition result of the recognition model whether the current baby is crying can be obtained, and the recognition result can be sent to the monitoring terminal, so that the guardian in the leaving state can timely see the reminder information. Because this application uses a variety of feature sequences to form audio feature vectors, the recognition results are more accurate, and through the existing smart device installation recognition application, you can effectively recognize baby crying, without the need to purchase additional special recognition equipment , Help to reduce system hardware costs.
  • FIG. 3 is a schematic diagram of an implementation process of yet another baby cry recognition method provided by an embodiment of the present application, which is described in detail as follows:
  • step S301 collect voice data and intercept voice data segments of a predetermined duration
  • step S302 calculate two or more audio feature vectors included in the voice data segment
  • Steps S301-S302 are basically the same as steps S201-S202 in FIG. 2.
  • step S303 it is determined whether the current network is in a connected state
  • the collection device when it is a device such as a smart phone, it may be in different network scenarios.
  • the smart device may interact with the cloud server through the WIFI network, or the collection device may be in a state of no network connection, but the collection device itself has a mobile communication module, such as a collection device
  • a phone card built in or the collection device can be connected to the network, and a phone card is built in. The way to send the results of these scenarios is discussed separately below.
  • step S304 if the current network is connected, the audio feature vector is sent to the cloud server, so that the cloud server sends an application reminder message to the monitoring terminal according to the recognition result.
  • the collection device can interact with the cloud server through the network, can send the collected audio data segment to the server, or calculate two or more audio feature vectors in the voice data segment ,
  • the audio feature vector is sent to the cloud server, and the cloud server recognizes the crying of the baby. If the cloud server recognizes that the audio data segment includes baby crying, it can send a prompt message to the monitoring terminal through the network, or it can also send a short message to the monitoring terminal, or make a network call.
  • the cloud server can also send the latest version number of the identification model to the collection terminal.
  • the identification model in the collection terminal is not the latest version, it can be sent to the cloud server Update request to download the latest recognition model.
  • step S305 if the current network is in a disconnected state, the audio feature vector is identified through a locally stored neural network model
  • the audio feature vector or voice data segment cannot be recognized by the server, and the audio feature vector can be locally recognized by storing the recognition model locally. Once the network is restored, the local recognition model can also be updated.
  • step S306 when the recognition result is a predetermined alarm result, a short message is sent to the monitoring terminal or an alarm call is made.
  • the recognition result is a predetermined alarm result, for example, when a baby cries are recognized, a short message is sent to the monitoring terminal or an alarm call is made to prompt the monitoring staff to take care of the baby in time.
  • the collection terminal may also collect the audio feature vector of the baby and receive the baby demand result input by the user to obtain a recognition model that can recognize the baby demand.
  • the baby's needs may include the needs for milk, warmth, cooling or safety.
  • the specific needs of the baby are output, and the specific needs are sent to the monitoring terminal, thereby improving the convenience of use of the guardian.
  • FIG. 4 is a schematic structural diagram of a baby cry recognition device according to an embodiment of the present application, which is described in detail as follows:
  • the baby cry recognition device includes:
  • the voice data collection unit 401 is used to collect voice data and intercept voice data segments of a predetermined duration
  • An audio feature vector calculation unit 402 configured to calculate two or more audio feature vectors included in the speech data segment
  • the recognition unit 403 is configured to recognize the audio feature vector according to a preset recognition model, and send the recognition result to the monitoring terminal.
  • the audio feature vector calculation unit includes:
  • a feature sequence calculation subunit configured to calculate two or more of a zero-crossing rate feature sequence, an energy feature sequence, a multi-order Mel frequency cepstrum coefficient feature sequence or a spectral centroid feature sequence in the voice data segment;
  • a selection subunit is used to select two or more feature sequences of a zero-crossing rate feature sequence, an energy feature sequence, a multi-order Mel frequency cepstrum coefficient feature sequence, or a spectral centroid feature sequence to generate an audio feature vector.
  • the baby cry recognition device shown in FIG. 4 corresponds to the baby cry recognition method shown in FIGS. 2-3.
  • FIG. 5 is a schematic diagram of a baby cry recognition device provided by an embodiment of the present application.
  • the baby cry recognition device 5 of this embodiment includes: a processor 50, a memory 51, and a computer program 52 stored in the memory 51 and executable on the processor 50, for example, a baby crying Sound recognition program.
  • the processor 50 executes the computer program 52, the steps in the foregoing embodiments of the baby cry recognition method are implemented.
  • the processor 50 executes the computer program 52, the functions of each module/unit in the foregoing device embodiments are realized.
  • the computer program 52 may be divided into one or more modules/units, and the one or more modules/units are stored in the memory 51 and executed by the processor 50 to complete This application.
  • the one or more modules/units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program 52 in the baby cry recognition device 5.
  • the computer program 52 may be divided into:
  • Voice data collection unit used to collect voice data and intercept voice data segments of a predetermined duration
  • An audio feature vector calculation unit configured to calculate two or more audio feature vectors included in the speech data segment
  • the recognition unit is configured to recognize the audio feature vector according to a preset recognition model, and send the recognition result to the monitoring terminal.
  • the baby cry recognition device may include, but is not limited to, the processor 50 and the memory 51.
  • FIG. 5 is only an example of the baby cry recognition device 5, and does not constitute a limitation on the baby cry recognition device 5, and may include more or fewer parts than shown, or a combination of certain Components, or different components, for example, the baby cry recognition device may further include an input and output device, a network access device, a bus, and the like.
  • the so-called processor 50 may be a central processing unit (Central Processing Unit (CPU), can also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (Application Specific Integrated Circuit (ASIC), ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • the memory 51 may be an internal storage unit of the baby cry recognition device 5, such as a hard disk or a memory of the baby cry recognition device 5.
  • the memory 51 may also be an external storage device of the baby cry recognition device 5, for example, a plug-in hard disk equipped on the baby cry recognition device 5, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, flash memory card (Flash Card) etc.
  • the memory 51 may also include both an internal storage unit of the baby cry recognition device 5 and an external storage device.
  • the memory 51 is used to store the computer program and other programs and data required by the baby cry recognition device.
  • the memory 51 can also be used to temporarily store data that has been or will be output.
  • the disclosed device/terminal device and method may be implemented in other ways.
  • the device/terminal device embodiments described above are only schematic.
  • the division of the module or unit is only a logical function division, and in actual implementation, there may be another division manner, such as multiple units Or components can be combined or integrated into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place or may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above integrated unit can be implemented in the form of hardware or software function unit.
  • the integrated module/unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium.
  • the present application can implement all or part of the processes in the methods of the above embodiments, or it can be completed by a computer program instructing related hardware.
  • the computer program can be stored in a computer-readable storage medium. When the program is executed by the processor, the steps of the foregoing method embodiments may be implemented.
  • the computer program includes computer program code, and the computer program code may be in a source code form, an object code form, an executable file, or some intermediate form, etc.
  • the computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signals, telecommunications signals, and software distribution media.
  • ROM Read-Only Memory
  • RAM Random Access Memory
  • electrical carrier signals telecommunications signals
  • software distribution media any entity or device capable of carrying the computer program code
  • a recording medium a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signals, telecommunications signals, and software distribution media.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Signal Processing (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Computational Linguistics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Telephonic Communication Services (AREA)

Abstract

一种婴儿哭声识别方法包括：采集语音数据，截取预定时长的语音数据段；计算所述语音数据段中包括的两个或者两个以上的音频特征向量；根据预先设定的识别模型识别所述音频特征向量，将所述识别结果发送至监护终端。由于识别结果是根据两个或两个以上的音频特征向量所识别，因而识别结果更加准确可靠，有利于提高婴儿哭声识别的精度。

Description

一种婴儿哭声识别方法、装置及设备 技术领域
本申请属于声音识别领域，尤其涉及一种婴儿哭声识别方法、装置及设备。
 背景技术
新生婴儿在表达情感需求或生理需求时，一般通过婴儿的哭声传递给外部世界。在现实生活中，由于照顾新生婴儿的工作基本托付给家里的老人或婴儿的母亲等监护人，并且监护人往往会同时身兼多职。因而，当婴儿处于睡眠状态的时候，监护人可能会忙于其它事情而处于离开状态。由于空间距离的原因，婴儿的啼哭并不一定能被监护人直接听到，使得监护人不能及时的听到婴儿的需求。
为了使得监护人能够及时的听到婴儿的需求，目前已经出现一些婴儿哭声的提醒装置，基于摄像头或穿戴式设备等数据采集设备，将数据采集设备接入网络，通过云端来完成识别工作。目前常见的识别方案是基于分贝、过零率或能量这些单一指标来确定是否需要报警，当环境中出现干扰音时，容易使得误报率较高。
 技术问题
有鉴于此，本申请实施例提供了一种婴儿哭声识别方法、装置及设备，以解决现有技术中识别婴儿哭声的方法中，当环境出现干扰时，容易使得误报率高的问题。
 技术解决方案
本申请实施例的第一方面提供了一种婴儿哭声识别方法，所述婴儿哭声识别方法包括：
采集语音数据，截取预定时长的语音数据段；
计算所述语音数据段中包括的两个或者两个以上的音频特征向量；
根据预先设定的识别模型识别所述音频特征向量，将所述识别结果发送至监护终端。
结合第一方面，在第一方面的第一种可能实现方式中，所述计算所述语音数据段中包括的两个或者两个以上的音频特征向量的步骤包括：
计算所述语音数据段中的过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者多种；
选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列生成音频特征向量。
结合第一方面的第一种可能实现方式，在第一方面的第二种可能实现方式中，所述选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列生成音频特征向量的步骤包括：
选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列，计算所选择的特征序列的均值；
根据所计算的均值确定所述音频特征向量。
结合第一方面，在第一方面的第三种可能实现方式中，所述根据预先设定的识别模型识别所述音频特征向量，将所述识别结果发送至监护终端的步骤包括：
判断当前网络是否处于连接状态；
如果当前网络处于连接状态，则将所述音频特征向量发送至云服务器，以使得云服务器根据识别结果向所述监护终端发送应用提醒消息。
结合第一方面的第三种可能实现方式，在第一方面的第四种可能实现方式中，所述方法还包括：
如果当前网络处于断开状态，则通过本地存储的神经网络模型识别所述音频特征向量；
当识别结果为预定的告警结果时，向监护终端发送短信息或者拨打告警电话。
结合第一方面，在第一方面的第五种可能实现方式中，在所述计算所述语音数据段中包括的两个或者两个以上的音频特征向量的步骤之前，所述方法还包括：
对所述语音数据段进行加重、分帧和加窗处理中的一项或者多项。
本申请实施例的第二方面提供了一种婴儿哭声识别装置，所述婴儿哭声识别装置包括：
语音数据采集单元，用于采集语音数据，截取预定时长的语音数据段；
音频特征向量计算单元，用于计算所述语音数据段中包括的两个或者两个以上的音频特征向量；
识别单元，用于根据预先设定的识别模型识别所述音频特征向量，将所述识别结果发送至监护终端。
结合第二方面，在第二方面的第一种可能实现方式中，所述音频特征向量计算单元包括：
特征序列计算子单元，用于计算所述语音数据段中的过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者多种；
选择子单元，用于选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列生成音频特征向量。
本申请实施例的第三方面提供了一种婴儿哭声识别设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如第一方面任一项所述婴儿哭声识别方法的步骤。
本申请实施例的第四方面提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时实现如第一方面任一项所述婴儿哭声识别方法的步骤。
 有益效果
本申请实施例与现有技术相比存在的有益效果是：通过采集语音数据，截取预定时长的语音数据段，计算所述语音数据段中包括的两个或两个以上的音频特征向量，根据预先设定的识别模型识别所述音频特征向量，并将识别结果发送至监护终端。由于识别结果是根据两个或两个以上的音频特征向量所识别，因而识别结果更加准确可靠，有利于提高婴儿哭声识别的精度。
 附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1是本申请实施例提供的一种婴儿哭声识别方法的实施场景示意图；
图2是本申请实施例提供的一种婴儿哭声识别方法的实现流程示意图；
图3是本申请实施例提供的又一婴儿哭声识别方法的实现流程示意图；
图4是本申请实施例提供的一种婴儿哭声识别装置的示意图；
图5是本申请实施例提供的婴儿哭声识别设备的示意图。
 本发明的实施方式
以下描述中，为了说明而不是为了限定，提出了诸如特定系统结构、技术之类的具体细节，以便透彻理解本申请实施例。然而，本领域的技术人员应当清楚，在没有这些具体细节的其它实施例中也可以实现本申请。在其它情况中，省略对众所周知的系统、装置、电路以及方法的详细说明，以免不必要的细节妨碍本申请的描述。
为了说明本申请所述的技术方案，下面通过具体实施例来进行说明。
图1为本申请实施例提供的一种婴儿哭声识别方法的实施场景示意图，如图1所示，所述实施场景包括采集终端、监护终端、云服务器。其中，所述采集终端可以为智能手机、平板电脑等。可以在采集终端安装应用程序的方式，免于额外配置硬件设备，有利于减少实现婴儿哭声识别的硬件成本。
所述采集终端用于采用婴儿的语音数据，根据预先设定的时长，对采集的语音数据进行分段截取。比如以当前时间为起点，截取当前时间之前的预定时长的一段时间内所采集的语音数据，得到语音数据段。所述预定时长可以为30秒等。在网络正常连接的时候，可以将所述语音数据段发送至云服务器，由所述云服务器对所采集的语音数据段进行计算，通过预先设定的识别模型，识别所述语音数据段中包括的两个或两个以上的音频特征向量，生成婴儿哭声识别结果。当然，在采集终端没有网络连接的情况下，也可以由采集终端本地存储的识别模型进行识别。当恢复网络连接的时候，可以通过识别模型版本号的比较，将云服务器中存储的最新的识别模型更新至所述采集终端。所述监护终端为监护人员携带的设备，可以为智能手机等。可以通过安装的应用程序接收提醒，或者通过短信或者电话的方式接收提醒。
图2为本申请实施例提供的一种婴儿哭声识别方法的实现流程示意图，详述如下：
在步骤S201中，采集语音数据，截取预定时长的语音数据段；
具体的，本申请所述婴儿哭声识别方法，可以基于现有的智能设备实现。在配备了麦克风的智能设备中安装本申请所述婴儿哭声识别方法对应的应用程序，可以有效的对采集的语音数据进行分析和处理，得到婴儿哭声的识别结果。
在采集语音数据后，可以根据预先设定的预定时长对语音数据进行分段截取。可以根据当前时间为语音数据段的结束时间，向当前时间之前取预定时长的语音数据，得到语音数据段。另外，所述语音数据段的截取时间间隔可以根据语音数据段的预定时长来设定。比如可以取语音数据段的预定时长的一定比例值。如预定时长为30秒时，所述语音数据段的截取时间可以为5秒等，从而能够对语音数据进行动态的分析。
在步骤S202中，计算所述语音数据段中包括的两个或者两个以上的音频特征向量；
所述语音数据段中包括的音频特征向量，可以包括过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者多种。其中，所述多阶梅尔频率倒谱系数特征序列可以为13阶梅尔频率倒谱系数特征序列。通过对音频的过零率、能量、多阶梅尔频率倒谱系数、频谱质心等特征中的两种或者两种以上的特征提取，得到融合了两个或者两个以上特征序列的音频特征向量。其中，音频的过零率，是指音频信号的符号变化的比率，符号变化包括如音频信号从正数变成负数，或者从负数变成正数。
所述能量可以为能量的大小变化的数值。
所述多阶梅尔频率倒谱，是基于声音频率的非线性梅尔刻度(mel scale)的对数能量频谱的线性变换。梅尔频率倒谱系数 (Mel-Frequency Cepstral Coefficients，MFCCs)就是组成梅尔频率倒谱的系数。它衍生自音讯片段的倒频谱(cepstrum)。
可以选择任意两个特征序列计算音频特征向量。当然，优选的实施方式为，选择包括过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列和频谱质心特征序列的多维特征序列。比如多阶梅尔频率倒谱系数特征序列为13阶时，则可以选择16维特征序列，从而便于能够得到更为精确的识别结果。
另外，所述音频特征向量，可以选择选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列直接得到音频特征向量。
或者，所述音频特征向量，也可以通过选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列，计算所选择的特征序列的均值，根据所计算的均值确定所述音频特征向量。
优选的一种实施方式中，在对所述语音数据段进行计算时，还可以包括对所述语音数据段进行加重、分帧和加窗处理中的一项或者多项的步骤。其中：
为了消除发声过程中，声带和嘴唇造成的效应，来补偿语音信号受到发音系统所压抑的高频部分，突显高频的共振峰，通过加重处理，在频域上乘以一个系数，这个系数可以跟频率成正相关，从而使得高频的幅值会有所提升。
虽然语音信号在宏观上不平稳，但是，在微观上具有短时平稳性（10---30ms内可以认为语音信号近似不变），根据微观的平衡性，可以把语音信号分为一些短段，即分帧来进行处理，每一个短段称为一帧。
为了便于进行傅里叶展开，还可以对语音数据段进行加窗处理，即将语音数据与一个窗函数相乘，使全局更加连续，避免出现吉布斯效应。通过加窗处理，使原本没有周期性的语音信号呈现出周期函数的部分特征。
另外，值得注意的是，由于加窗处理会使得一帧信号的两端被削弱，因此，在分帧的时候，帧与帧之间需要有重叠。
在步骤S203中，根据预先设定的识别模型识别所述音频特征向量，将所述识别结果发送至监护终端。
所述识别模型可以为神经网络模型。可以采集大量的婴儿哭声样本和噪声样本，计算哭声样本和噪声样本的音频特征向量，对所述神经网络模型进行训练，根据训练完成的神经网络模型，对所述音频特征向量进行识别处理。可选的一种实施方式中，可以采集被监护婴儿的哭声数据，对所述识别模型进行训练，以使得能够得到更为可靠的识别结果。
根据识别模型的识别结果，可以得到当前婴儿是否有婴儿哭声，可以将所述识别结果发送至监护终端，从而使得处于离开状态的监护人能够及时的看到提醒信息。由于本申请采用多种特征序列构成音频特征向量，从而使得识别结果更加准确，并且通过现有的智能设备安装识别应用程序，即可有效的进行婴儿哭声识别，不需要另外购置专门的识别设备，有利于减少系统硬件成本。
图3为本申请实施例提供的又一婴儿哭声识别方法的实现流程示意图，详述如下：
在步骤S301中，采集语音数据，截取预定时长的语音数据段；
在步骤S302中，计算所述语音数据段中包括的两个或者两个以上的音频特征向量；
步骤S301-S302与图2中的步骤S201-S202基本相同。
在步骤S303中，判断当前网络是否处于连接状态；
在本申请中，所述采集设备为智能手机等设备时，可能会处于不同的网络场景。比如，可能采集设备处于有WIFI网络的场景中，智能设备可以通过所述WIFI网络与云服务器交互，或者，采集设备可能处于无网络连接的状态，但采集设备本身具有移动通信模块，比如采集设备中内置有电话卡，或者，采集设备即可以连接网络，又内置有电话卡，下面对这些场景的结果发送方式分别讨论。
在步骤S304中，如果当前网络处于连接状态，则将所述音频特征向量发送至云服务器，以使得云服务器根据识别结果向所述监护终端发送应用提醒消息。
当采集设备的网络处于连接状态时，采集设备可以通过网络与云服务器进行交互，可以将采集的音频数据段发送至服务器，或者计算语音数据段中的两个或两个以上的音频特征向量后，将音频特征向量发送至云服务器，由云服务器进行婴儿哭声的识别。如果云服务器识别音频数据段中包括婴儿哭声时，则可以通过网络向监护终端发送提示消息，或者也可以向监护终端发送短信息，或者拨打网络电话等。
当然，在网络处于连接状态时，所述云服务器还可以将识别模型的最新版本号发送至采集终端，采集终端通过比较，如果采集终端中的识别模型不是最新版本时，则可以向云服务器发送更新请求，下载最新的识别模型。
另外，在步骤S305中，如果当前网络处于断开状态，则通过本地存储的神经网络模型识别所述音频特征向量；
如果当前网络处于断开状态，则无法由服务器对音频特征向量或语音数据段进行识别，可以通过在本地存储识别模型的方式，由本地对所述音频特征向量进行识别。一旦恢复网络，还可以更新本地所存储的识别模型。
在步骤S306中，当识别结果为预定的告警结果时，向监护终端发送短信息或者拨打告警电话。
如果识别结果为预定的告警结果，比如识别到婴儿哭声时，则向监护终端发送短消息或者拨打告警电话，提示监护人员及时照看婴儿。
当然，作为本申请优化的一种实施方式，也可以由采集终端采集婴儿的音频特征向量，并接收用户输入的婴儿需求结果，得到能够识别婴儿需求的识别模型。所述婴儿需求可以包括吃奶需求、保暖需求、降温需求或安全感需求等。将采集的音频特征向量输入所述识别模型时，输出婴儿的具体需求，并将所述具体需求发送至监护终端，提高监护人的使用便利性。
应理解，上述实施例中各步骤的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
图4为本申请实施例提供的一种婴儿哭声识别装置的结构示意图，详述如下：
所述婴儿哭声识别装置包括：
语音数据采集单元401，用于采集语音数据，截取预定时长的语音数据段；
音频特征向量计算单元402，用于计算所述语音数据段中包括的两个或者两个以上的音频特征向量；
识别单元403，用于根据预先设定的识别模型识别所述音频特征向量，将所述识别结果发送至监护终端。
优选的，所述音频特征向量计算单元包括：
特征序列计算子单元，用于计算所述语音数据段中的过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者多种；
选择子单元，用于选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列生成音频特征向量。
图4所述婴儿哭声识别装置，与图2-3所述的婴儿哭声识别方法对应。
图5是本申请一实施例提供的婴儿哭声识别设备的示意图。如图5所示，该实施例的婴儿哭声识别设备5包括：处理器50、存储器51以及存储在所述存储器51中并可在所述处理器50上运行的计算机程序52，例如婴儿哭声识别程序。所述处理器50执行所述计算机程序52时实现上述各个婴儿哭声识别方法实施例中的步骤。或者，所述处理器50执行所述计算机程序52时实现上述各装置实施例中各模块/单元的功能。
示例性的，所述计算机程序52可以被分割成一个或多个模块/单元，所述一个或者多个模块/单元被存储在所述存储器51中，并由所述处理器50执行，以完成本申请。所述一个或多个模块/单元可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述所述计算机程序52在所述婴儿哭声识别设备5中的执行过程。例如，所述计算机程序52可以被分割成：
语音数据采集单元，用于采集语音数据，截取预定时长的语音数据段；
音频特征向量计算单元，用于计算所述语音数据段中包括的两个或者两个以上的音频特征向量；
识别单元，用于根据预先设定的识别模型识别所述音频特征向量，将所述识别结果发送至监护终端。
所述婴儿哭声识别设备可包括，但不仅限于，处理器50、存储器51。本领域技术人员可以理解，图5仅仅是婴儿哭声识别设备5的示例，并不构成对婴儿哭声识别设备5的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如所述婴儿哭声识别设备还可以包括输入输出设备、网络接入设备、总线等。
所称处理器50可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器 (Digital Signal Processor，DSP)、专用集成电路 (Application Specific Integrated Circuit，ASIC)、现成可编程门阵列 (Field-Programmable Gate Array，FPGA) 或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
所述存储器51可以是所述婴儿哭声识别设备5的内部存储单元，例如婴儿哭声识别设备5的硬盘或内存。所述存储器51也可以是所述婴儿哭声识别设备5的外部存储设备，例如所述婴儿哭声识别设备5上配备的插接式硬盘，智能存储卡（Smart Media Card, SMC），安全数字（Secure Digital, SD）卡，闪存卡（Flash Card）等。进一步地，所述存储器51还可以既包括所述婴儿哭声识别设备5的内部存储单元也包括外部存储设备。所述存储器51用于存储所述计算机程序以及所述婴儿哭声识别设备所需的其他程序和数据。所述存储器51还可以用于暂时地存储已经输出或者将要输出的数据。
所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，仅以上述各功能单元、模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能单元、模块完成，即将所述装置的内部结构划分成不同的功能单元或模块，以完成以上描述的全部或者部分功能。实施例中的各功能单元、模块可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中，上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。另外，各功能单元、模块的具体名称也只是为了便于相互区分，并不用于限制本申请的保护范围。上述系统中单元、模块的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述或记载的部分，可以参见其它实施例的相关描述。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
在本申请所提供的实施例中，应该理解到，所揭露的装置/终端设备和方法，可以通过其它的方式实现。例如，以上所描述的装置/终端设备实施例仅仅是示意性的，例如，所述模块或单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通讯连接可以是通过一些接口，装置或单元的间接耦合或通讯连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的模块/单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请实现上述实施例方法中的全部或部分流程，也可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一计算机可读存储介质中，该计算机程序在被处理器执行时，可实现上述各个方法实施例的步骤。。其中，所述计算机程序包括计算机程序代码，所述计算机程序代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。所述计算机可读介质可以包括：能够携带所述计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器（ROM，Read-Only Memory）、随机存取存储器（RAM，Random Access Memory）、电载波信号、电信信号以及软件分发介质等。需要说明的是，所述计算机可读介质包含的内容可以根据司法管辖区内立法和专利实践的要求进行适当的增减，例如在某些司法管辖区，根据立法和专利实践，计算机可读介质不包括是电载波信号和电信信号。
以上所述实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围，均应包含在本申请的保护范围之内。

Claims (10)

  1. 一种婴儿哭声识别方法，其特征在于，所述婴儿哭声识别方法包括：
    采集语音数据，截取预定时长的语音数据段；
    计算所述语音数据段中包括的两个或者两个以上的音频特征向量；
    根据预先设定的识别模型识别所述音频特征向量，将所述识别结果发送至监护终端。
  2. 根据权利要求1所述的婴儿哭声识别方法，其特征在于，所述计算所述语音数据段中包括的两个或者两个以上的音频特征向量的步骤包括：
    计算所述语音数据段中的过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者多种；
    选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列生成音频特征向量。
  3. 根据权利要求2所述的婴儿哭声识别方法，其特征在于，所述选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列生成音频特征向量的步骤包括：
    选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列，计算所选择的特征序列的均值；
    根据所计算的均值确定所述音频特征向量。
  4. 根据权利要求1所述的婴儿哭声识别方法，其特征在于，所述根据预先设定的识别模型识别所述音频特征向量，将所述识别结果发送至监护终端的步骤包括：
    判断当前网络是否处于连接状态；
    如果当前网络处于连接状态，则将所述音频特征向量发送至云服务器，以使得云服务器根据识别结果向所述监护终端发送应用提醒消息。
  5. 根据权利要求4所述的婴儿哭声识别方法，其特征在于，所述方法还包括：
    如果当前网络处于断开状态，则通过本地存储的神经网络模型识别所述音频特征向量；
    当识别结果为预定的告警结果时，向监护终端发送短信息或者拨打告警电话。
  6. 根据权利要求1所述的婴儿哭声识别方法，其特征在于，在所述计算所述语音数据段中包括的两个或者两个以上的音频特征向量的步骤之前，所述方法还包括：
    对所述语音数据段进行加重、分帧和加窗处理中的一项或者多项。
  7. 一种婴儿哭声识别装置，其特征在于，所述婴儿哭声识别装置包括：
    语音数据采集单元，用于采集语音数据，截取预定时长的语音数据段；
    音频特征向量计算单元，用于计算所述语音数据段中包括的两个或者两个以上的音频特征向量；
    识别单元，用于根据预先设定的识别模型识别所述音频特征向量，将所述识别结果发送至监护终端。
  8. 根据权利要求7所述的婴儿哭声识别装置，其特征在于，所述音频特征向量计算单元包括：
    特征序列计算子单元，用于计算所述语音数据段中的过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者多种；
    选择子单元，用于选择过零率特征序列、能量特征序列、多阶梅尔频率倒谱系数特征序列或频谱质心特征序列中的两种或者两种以上的特征序列生成音频特征向量。
  9. 一种婴儿哭声识别设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，其特征在于，所述处理器执行所述计算机程序时实现如权利要求1至6任一项所述婴儿哭声识别方法的步骤。
  10. 一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现如权利要求1至6任一项所述婴儿哭声识别方法的步骤。
PCT/CN2019/130824 2019-01-12 2019-12-31 一种婴儿哭声识别方法、装置及设备 WO2020143512A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910029052.4 2019-01-12
CN201910029052.4A CN109658953A (zh) 2019-01-12 2019-01-12 一种婴儿哭声识别方法、装置及设备

Publications (1)

Publication Number Publication Date
WO2020143512A1 true WO2020143512A1 (zh) 2020-07-16

Family

ID=66119244

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/130824 WO2020143512A1 (zh) 2019-01-12 2019-12-31 一种婴儿哭声识别方法、装置及设备

Country Status (2)

Country Link
CN (1) CN109658953A (zh)
WO (1) WO2020143512A1 (zh)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109658953A (zh) * 2019-01-12 2019-04-19 深圳先进技术研究院 一种婴儿哭声识别方法、装置及设备
CN110415724A (zh) * 2019-08-08 2019-11-05 中南大学湘雅二医院 报警数据的传输方法、装置、系统及计算机可读存储介质
CN111128100B (zh) * 2019-12-20 2021-04-20 网易（杭州）网络有限公司 节奏点检测方法、装置及电子设备
CN111370025A (zh) * 2020-02-25 2020-07-03 广州酷狗计算机科技有限公司 音频识别方法、装置及计算机存储介质
EP3940698A1 (en) 2020-07-13 2022-01-19 Zoundream AG A computer-implemented method of providing data for an automated baby cry assessment
CN112270932A (zh) * 2020-10-22 2021-01-26 北京小米松果电子有限公司 用于智能设备的报警方法及装置、电子设备及存储介质
CN112382302A (zh) * 2020-12-02 2021-02-19 漳州立达信光电子科技有限公司 婴儿哭声识别方法及终端设备
CN112992136A (zh) * 2020-12-16 2021-06-18 呼唤(上海)云计算股份有限公司 智能婴儿监护系统及方法
CN113436650B (zh) * 2021-08-25 2021-11-16 深圳市北科瑞声科技股份有限公司 婴儿哭声识别方法、装置、电子设备及存储介质
CN116825115A (zh) * 2023-08-10 2023-09-29 广州番禺职业技术学院 一种婴儿哭声识别方法、系统及装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2010210730A (ja) * 2009-03-09 2010-09-24 Univ Of Fukui 乳幼児の感情診断装置及び方法
CN106653001A (zh) * 2016-11-17 2017-05-10 沈晓明 婴儿哭声的辩识方法及其系统
CN107808658A (zh) * 2016-09-06 2018-03-16 深圳声联网科技有限公司 基于家居环境下实时的婴儿音频系列行为检测方法
CN107818779A (zh) * 2017-09-15 2018-03-20 北京理工大学 一种婴幼儿啼哭声检测方法、装置、设备及介质
CN109658953A (zh) * 2019-01-12 2019-04-19 深圳先进技术研究院 一种婴儿哭声识别方法、装置及设备

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2466242B (en) * 2008-12-15 2013-01-02 Audio Analytic Ltd Sound identification systems
CN102163427B (zh) * 2010-12-20 2012-09-12 北京邮电大学 一种基于环境模型的音频异常事件检测方法
KR20120107382A (ko) * 2011-03-21 2012-10-02 호서대학교 산학협력단 유아 울음 분석 장치
CN103280220B (zh) * 2013-04-25 2015-11-18 北京大学深圳研究生院 一种实时的婴儿啼哭声识别方法
CN103489282B (zh) * 2013-09-24 2016-01-20 华南理工大学 一种能识别婴儿哭声的婴儿监护器及婴儿哭声识别方法
BR112017012604A2 (pt) * 2014-12-16 2018-01-16 Koninklijke Philips Nv monitor de sono, método de monitoramento automático de sono de bebê, e produto de programa de computador

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2010210730A (ja) * 2009-03-09 2010-09-24 Univ Of Fukui 乳幼児の感情診断装置及び方法
CN107808658A (zh) * 2016-09-06 2018-03-16 深圳声联网科技有限公司 基于家居环境下实时的婴儿音频系列行为检测方法
CN106653001A (zh) * 2016-11-17 2017-05-10 沈晓明 婴儿哭声的辩识方法及其系统
CN107818779A (zh) * 2017-09-15 2018-03-20 北京理工大学 一种婴幼儿啼哭声检测方法、装置、设备及介质
CN109658953A (zh) * 2019-01-12 2019-04-19 深圳先进技术研究院 一种婴儿哭声识别方法、装置及设备

Also Published As

Publication number Publication date
CN109658953A (zh) 2019-04-19

Similar Documents

Publication Publication Date Title
WO2020143512A1 (zh) 一种婴儿哭声识别方法、装置及设备
CN107799126B (zh) 基于有监督机器学习的语音端点检测方法及装置
US9609441B2 (en) Smart hearing aid
CN111667818B (zh) 一种训练唤醒模型的方法及装置
US20200228907A1 (en) Mobile device based techniques for detection and prevention of hearing loss
CN111770427A (zh) 麦克风阵列的检测方法、装置、设备以及存储介质
CN108335694A (zh) 远场环境噪声处理方法、装置、设备和存储介质
CN110444202B (zh) 复合语音识别方法、装置、设备及计算机可读存储介质
WO2019242402A1 (zh) 语音识别模型生成方法、装置、存储介质及电子设备
US11133022B2 (en) Method and device for audio recognition using sample audio and a voting matrix
CN111868823A (zh) 一种声源分离方法、装置及设备
WO2023103253A1 (zh) 一种音频检测方法、装置及终端设备
CN113239872A (zh) 事件识别方法、装置、设备及存储介质
CN104282303B (zh) 利用声纹识别进行语音辨识的方法及其电子装置
CN112382302A (zh) 婴儿哭声识别方法及终端设备
CN115426582A (zh) 一种耳机音频处理方法及装置
CN111339881A (zh) 基于情绪识别的宝宝成长监护方法及系统
CN112259076B (zh) 语音交互方法、装置、电子设备及计算机可读存储介质
US11908464B2 (en) Electronic device and method for controlling same
CN109767751B (zh) 婴儿安抚音乐的制作方法、装置、计算机设备及存储装置
CN112306560B (zh) 用于唤醒电子设备的方法和装置
CN111916107A (zh) 一种音频分类模型的训练方法、音频的分类方法及装置
CN114495976A (zh) 语音测试方法、装置、系统、电子设备及存储介质
CN113889086A (zh) 语音识别模型的训练方法、语音识别方法及相关装置
CN109975795B (zh) 一种声源追踪方法和装置

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: 19909073

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: 19909073

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 19909073

Country of ref document: EP

Kind code of ref document: A1