Focusing type voiceprint acquisition sensor

Abstract

The invention relates to a focusing type voiceprint acquisition sensor, and discloses a sensor which realizes accurate focusing of equipment operation sound, reduction of environmental noise interference, improvement of sound signal quality, multi-dimensional voiceprint feature extraction and convenience for subsequent equipment operation state analysis through cooperation of an array type multi-direction sound acquisition module and a focusing concave mirror. The device is characterized in that the device is composed of a sound insulation cover, a sound collection hole, a voiceprint collection cylinder, a multi-angle voiceprint collection cylinder, a connecting rod, a sound collection module, a voice processing module and a focusing concave mirror, one end of the voiceprint collection cylinder is arranged on the multi-angle voiceprint collection cylinder and communicates with the multi-angle voiceprint collection cylinder, and the other end of the voiceprint collection cylinder is of an opening structure; the sound insulation cover is arranged at the other end of the voiceprint collection cylinder, sound collection holes which are densely arranged are formed in the sound insulation cover, the sound collection holes are tiny holes, the sound collection holes are close to the upper edge of the sound insulation cover, and the focusing concave mirror is arranged in the voiceprint collection cylinder and corresponds to the sound collection holes.

Description

Technical Field

[0001] The present invention discloses a focused voiceprint collection sensor, which relates to a voiceprint collection sensor capable of focusing on a target sound, and belongs to the technical field of sound collection and analysis, and in particular to a sensor that uses an array-type multi-directional sound collection module in conjunction with a focusing concave mirror to achieve precise focusing on the sound of equipment operation, reduce environmental noise interference, improve sound signal quality, and extract multi-dimensional voiceprint features, so as to facilitate subsequent equipment operation status analysis. Background Art

[0002] With the continuous improvement of industrial automation, equipment such as motors and transformers play an important role in power systems, manufacturing and other industrial fields. Once these equipment fail, it will not only cause production interruption or unstable power supply, but also may cause safety hazards and even cause significant economic losses. Therefore, how to achieve equipment fault warning through effective means, timely discover potential problems and take preventive measures is one of the current research hotspots in the industrial field. During operation, the equipment will emit specific sound signals, which are called "voiceprints". The voiceprint characteristics of the equipment can reflect the changes in its operating status: during normal operation, the equipment has stable frequency components and regular vibration modes; when the equipment fails (such as bearing wear, winding short circuit, etc.), the voiceprint characteristics will change significantly. By analyzing the voiceprint characteristics of the equipment, the health status of the equipment can be evaluated and a reliable basis for fault warning can be provided. At present, equipment fault detection technology based on voiceprint characteristics has gradually been applied to practice, and voiceprint acquisition technology usually uses ordinary microphones or other sound sensors to capture sound signals when the equipment is running. Due to the complex background noise in the industrial environment (such as mechanical vibration, electromagnetic interference, etc.), traditional voiceprint acquisition devices are difficult to effectively distinguish the target sound from the environmental noise, resulting in a low signal-to-noise ratio of the collected sound signal, and only a single or a small number of microphones are used for sound acquisition, which cannot effectively cover the multi-directional sound inside the equipment, and it is difficult to fully capture the sound changes when the equipment is running, which directly affects the subsequent voiceprint feature extraction and fault warning analysis effect.

[0003] Publication No. CN216749300U discloses a voiceprint collection system, comprising: a processor, a voiceprint collection module electrically connected to the processor, and a display module; wherein the voiceprint collection module is suitable for collecting sound sources, and sending them to the processor after preliminary processing, that is, the processor receives the preliminary processing signal and performs deep processing on the preliminary processing signal; and the processor displays the processing result through the display module; Publication No. CN221127489U discloses a strong magnetic patch-type voiceprint collection device, comprising a patch-type voiceprint collection device body, the patch-type voiceprint collection device body comprises a patch-type voiceprint collection sensor, and the bottom of the patch-type voiceprint collection sensor is connected to a strong magnetic adsorption component, the outer ring of the patch-type voiceprint collection sensor is connected to a support mounting component for positioning the strong magnetic adsorption component, and the support mounting component comprises an integrally connected vertical block portion and a horizontal block portion. The above-mentioned voiceprint collection device uses a sound sensor to capture the sound signal when the equipment is running. Due to the complex background noise in the industrial environment (such as mechanical vibration, electromagnetic interference, etc.), it is difficult for traditional voiceprint collection devices to effectively distinguish the target sound from the environmental noise, resulting in a low signal-to-noise ratio of the collected sound signal. In addition, only a single or a small number of sound sensors are used for sound collection, which cannot achieve effective coverage of multi-directional sounds inside the equipment, and it is difficult to fully capture the sound changes when the equipment is running, which directly affects the subsequent voiceprint feature extraction and fault warning analysis effects. Summary of the invention

[0004] In order to improve the above situation, the present invention provides a focused voiceprint collection sensor that uses an array-type multi-directional sound collection module in conjunction with a focusing concave mirror to achieve precise focusing of the equipment operation sound, reduce environmental noise interference, improve sound signal quality, and extract multi-dimensional voiceprint features, which is convenient for subsequent equipment operation status analysis.

[0005] A focused voiceprint collection sensor of the present invention is implemented as follows: A focused voiceprint collection sensor of the present invention is composed of a sound insulation cover, a sound collection hole, a voiceprint collection tube, a multi-angle voiceprint collection tube, a connecting rod, a sound collection module, a voice processing module and a focusing concave mirror. One end of the voiceprint collection tube is placed on the multi-angle voiceprint collection tube and is connected to the multi-angle voiceprint collection tube. Preferably, the voiceprint collection tube and the multi-angle voiceprint collection tube are detachably connected. Preferably, the surface of the voiceprint collection tube and the multi-angle voiceprint collection tube is coated with three layers of sound-absorbing layers, and the three layers of sound-absorbing layers absorb high-frequency, medium-frequency and low-frequency sounds respectively. Preferably, the inner walls of the voiceprint collection tube and the multi-angle voiceprint collection tube are made of porous sound-absorbing material. Preferably, there are multiple voiceprint collection tubes, and the multiple voiceprint collection tubes are arranged equidistantly along the circumference of the multi-angle voiceprint collection tube to form an array-type multi-directional sound collection. Preferably, two adjacent voiceprint collection tubes are not on the same horizontal plane. The other end of the voiceprint collection tube is an open structure. A soundproof cover is placed on the other end of the voiceprint collection tube. The soundproof cover is provided with densely arranged sound collection holes. The sound collection holes are tiny holes. The sound collection holes are close to the upper edge of the soundproof cover. Preferably, a sound amplification module is provided on the inner wall of the sound insulation cover at a position corresponding to the sound collection hole. Preferably, the sound insulation cover has a multi-layer composite structure, the innermost layer is made of high-density sound-absorbing cotton, the middle layer is a damping material, and the outer layer is insulating ceramic. Preferably, the sound insulation cover is a concave structure. Preferably, the sound collection holes are arranged in a hexagonal honeycomb pattern, with a hole diameter between 1 mm and 3 mm. Preferably, the sound collecting hole is a tapered hole, which gradually narrows from the hole opening to the hole end. The focusing concave mirror is placed in the voiceprint collection tube and corresponds to the sound collection hole. The sound collection module is placed in the voiceprint collection tube. Preferably, the sound collection module is placed in the voiceprint collection tube via a connecting rod, and the connecting rod is a telescopic rod with adjustable length. Preferably, the sound collection module is located between the focusing concave mirror and the sound collection hole, and the heights of the focusing concave mirror and the sound collection hole are higher than the sound collection module. The sound of the device is transmitted to the focusing concave mirror through the sound collection hole, and then forms a reflection focus on the concave arc surface of the focusing concave mirror and falls on the sound collection module. Preferably, the focusing concave mirror surface is designed in a parabolic shape, and the reflectivity is above 95%. Preferably, a sound absorbing layer is provided on the back of the focusing concave mirror. The voice processing module is placed in the multi-angle voiceprint collection tube, and the sound collection module transmits signals to the voice processing module through the signal line. The voice processing module is provided with a digital filter. After receiving the sound signal collected by the sound collection module, the voice processing module performs voiceprint feature extraction. The voiceprint feature extraction method comprises the following steps: (1) Remove background noise and high-frequency noise through digital filters; (2) Normalize the amplitude of the sound signal to eliminate the impact of differences in sound intensity between different sensors or under different environmental conditions; (3) The continuous sound signal is divided into multiple time windows (i.e. frames), with the length of each frame ranging from 20 to 100 ms, and the Hanning window method is used to reduce inter-frame interference; (4) Perform fast Fourier transform (FFT) on each frame of sound signal to convert the time domain signal into a frequency domain signal; (5) Calculate the power spectral density (PSD) of the sound signal and draw a spectrum diagram to obtain the main frequency components and their energy distribution in the signal; (6) Extract the main frequency peaks and their corresponding amplitude and phase information from the spectrum diagram; (7) Calculate the mean and variance of the sound signal in each time window to characterize the fluctuation intensity of the signal; (8) Analyze the sharpness and asymmetry of the probability density function (PDF) of the sound signal and extract the kurtosis and skewness features; (9) Integrate the frequency component characteristics, energy distribution characteristics, fluctuation characteristics, and kurtosis and skewness characteristics to construct a multi-dimensional voiceprint feature vector to comprehensively characterize the operating status of the device; Furthermore, an insulating film is provided on the surface of the multi-angle voiceprint collection tube and the voiceprint collection tube, and the insulating film completely wraps the multi-angle voiceprint collection tube and the voiceprint collection tube; Furthermore, a collection sealing sleeve is placed between the sound insulation cover and the voiceprint collection tube. The collection seal is made of flexible material, which can reduce sound leakage, prevent external noise from entering the device from the joint, and improve the overall noise reduction effect. Beneficial Effects

[0006] 1. Through sound focusing, the quality of sound collection inside the device can be significantly improved and the impact of environmental noise can be reduced.

[0007] 2. Through the array-type multi-directional and multi-angle voiceprint collection tube, the three-dimensional and multi-angle coverage collection of the equipment operation sound is achieved.

[0008] 3. Constructing a multi-dimensional voiceprint feature vector that includes frequency component characteristics, energy distribution characteristics, fluctuation characteristics, as well as kurtosis and skewness characteristics can more comprehensively characterize the operating status of the device.

[0009] 4. Multi-level sound insulation design, with good noise reduction effect and insulation performance, suitable for soundprint collection of live equipment such as motors and transformers. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A three-dimensional structural diagram of a focused voiceprint acquisition sensor of the present invention; Figure 2 This is a structural schematic diagram of a focused voiceprint collection sensor of the present invention; Figure 3 This is a structural schematic diagram of a focused voiceprint collection sensor of the present invention; Figure 4 It is a three-dimensional structural diagram of embodiment 2 of a focused voiceprint acquisition sensor of the present invention; Figure 5 This is a three-dimensional structural diagram of Example 3 of a focused voiceprint acquisition sensor of the present invention. Attached photos

[0011] The components include: a sound insulation cover (1), a sound collection hole (2), a voiceprint collection tube (3), a multi-angle voiceprint collection tube (4), a connecting rod (5), a sound collection module (6), a voice processing module (7), a focusing concave mirror (8), an insulating film (9), and a collection sealing sleeve (10). DETAILED DESCRIPTION Example 1

[0012] A focusing voiceprint collection sensor of the present invention comprises a sound insulation cover (1), a sound collection hole (2), a voiceprint collection tube (3), a multi-angle voiceprint collection tube (4), a connecting rod (5), a sound collection module (6), a voice processing module (7) and a focusing concave mirror (8). One end of the voiceprint collection tube (3) is placed on the multi-angle voiceprint collection tube (4) and is connected to the multi-angle voiceprint collection tube (4). Preferably, the voiceprint collection tube (3) and the multi-angle voiceprint collection tube (4) are detachably connected. Preferably, the surfaces of the voiceprint collection tube (3) and the multi-angle voiceprint collection tube (4) are coated with three layers of sound-absorbing layers, and the three layers of sound-absorbing layers absorb high-frequency, medium-frequency and low-frequency sounds respectively. Preferably, the inner walls of the voiceprint collection tube (3) and the multi-angle voiceprint collection tube (4) are made of porous sound-absorbing material. Preferably, there are a plurality of the voiceprint collection tubes (3), and the plurality of voiceprint collection tubes (3) are arranged equidistantly along the circumference of the multi-angle voiceprint collection tube (4) to form an array-type multi-directional sound collection. Preferably, two adjacent voiceprint collection tubes (3) are not on the same horizontal plane. The other end of the voiceprint collection tube (3) is an open structure. The soundproof cover (1) is placed on the other end of the voiceprint collection tube (3). The soundproof cover (1) is provided with densely arranged sound collection holes (2). The sound collection holes (2) are tiny holes. The sound collection holes (2) are close to the upper edge of the soundproof cover (1). Preferably, a sound amplification module is provided on the inner wall of the soundproof cover (1) at a position corresponding to the sound collection hole (2). Preferably, the sound insulation cover (1) has a multi-layer composite structure, the innermost layer is made of high-density sound-absorbing cotton, the middle layer is made of damping material, and the outer layer is made of insulating ceramic. Preferably, the sound insulation cover (1) is a concave structure. Preferably, the sound collecting holes (2) are arranged in a hexagonal honeycomb pattern, with a hole diameter between 1 mm and 3 mm. Preferably, the sound collecting hole (2) is a tapered hole, which gradually narrows from the hole opening to the hole end. The focusing concave mirror (8) is placed in the voiceprint collection tube (3) and corresponds to the sound collection hole (2). The sound collection module (6) is placed in the voiceprint collection tube (3). Preferably, the sound collection module (6) is placed in the voiceprint collection tube (3) via a connecting rod (5), and the connecting rod (5) is a telescopic rod with adjustable length. Preferably, the sound collecting module (6) is located between the focusing concave mirror (8) and the sound collecting hole (2), and the heights of the focusing concave mirror (8) and the sound collecting hole (2) are higher than the sound collecting module (6). The operating sound of the device is transmitted to the focusing concave mirror (8) through the sound collection hole (2), and forms a reflection focus on the concave arc surface of the focusing concave mirror (8) and falls on the sound collection module (6). Preferably, the surface of the focusing concave mirror (8) is designed in a parabolic shape, and the reflectivity is above 95%. Preferably, a sound absorbing layer is provided on the back of the focusing concave mirror (8). The voice processing module (7) is placed in the multi-angle voiceprint collection tube (4), and the sound collection module (6) transmits signals to the voice processing module (7) via a signal line. The voice processing module (7) is provided with a digital filter. After receiving the sound signal collected by the sound collection module (6), the voice processing module (7) performs voiceprint feature extraction. The voiceprint feature extraction method comprises the following steps: (1) Remove background noise and high-frequency noise through digital filters; (2) Normalize the amplitude of the sound signal to eliminate the impact of differences in sound intensity between different sensors or under different environmental conditions; (3) The continuous sound signal is divided into multiple time windows (i.e. frames), with the length of each frame ranging from 20 to 100 ms, and the Hanning window method is used to reduce inter-frame interference; (4) Perform fast Fourier transform (FFT) on each frame of sound signal to convert the time domain signal into a frequency domain signal; (5) Calculate the power spectral density (PSD) of the sound signal and draw a spectrum diagram to obtain the main frequency components and their energy distribution in the signal; (6) Extract the main frequency peaks and their corresponding amplitude and phase information from the spectrum diagram; (7) Calculate the mean and variance of the sound signal in each time window to characterize the fluctuation intensity of the signal; (8) Analyze the sharpness and asymmetry of the probability density function (PDF) of the sound signal and extract the kurtosis and skewness features; (9) Integrate the frequency component characteristics, energy distribution characteristics, fluctuation characteristics, and kurtosis and skewness characteristics to construct a multi-dimensional voiceprint feature vector to comprehensively characterize the operating status of the device; When in use, the sensor is installed inside a device such as a motor or a transformer, and the plurality of voiceprint collection tubes (3) face different directions. The operating sounds at different positions inside the device are transmitted to the focusing concave mirror (8) through the sound collection hole (2). The sound waves are refracted by the focusing concave mirror (8) and focused to the sound collection module (6). The sound collection module (6) transmits the focused high-quality sound to the voice processing module (7), and the voice processing module (7) extracts the voiceprint of the sound. Example 2

[0013] The difference between this embodiment and embodiment 1 is that an insulating film (9) is provided on the surface of the multi-angle voiceprint collection tube (4) and the voiceprint collection tube (3), and the insulating film (9) completely wraps the multi-angle voiceprint collection tube (4) and the voiceprint collection tube (3); when in use, the multi-angle voiceprint collection tube (4) and the voiceprint collection tube (3) can be effectively prevented from contacting with external conductive materials, thereby avoiding the risk of short circuit or leakage, and improving the safety of the sensor in electromagnetic equipment such as motors and transformers; Example 3

[0014] The difference between this embodiment and embodiment 1 is that a collection sealing sleeve (10) is placed between the soundproof cover (1) and the voiceprint collection tube (3); the collection sealing sleeve (10) is made of a flexible material and can reduce sound leakage when in use, prevent external noise from entering the device from the joint, and improve the overall noise reduction effect; The surfaces of the voiceprint collection tube (3) and the multi-angle voiceprint collection tube (4) are coated with three layers of sound-absorbing layers. The three layers of sound-absorbing layers are designed to absorb high-frequency, medium-frequency and low-frequency sounds respectively, and can absorb noises of different frequencies in a targeted manner, thereby reducing the interference of environmental noise on sound signals; The inner walls of the voiceprint collection tube (3) and the multi-angle voiceprint collection tube (4) are designed with porous sound-absorbing materials, which can effectively absorb residual noise and reverberation sound, and reduce the interference of environmental noise on sound signals; The design of two adjacent voiceprint collection tubes (3) not being on the same horizontal plane can cover a wider range, improve the ability to capture the sound signal of the equipment operation, and make the sound collection more comprehensive; The sound insulation cover (1) has a multi-layer composite structure, with the innermost layer made of high-density sound-absorbing cotton, the middle layer made of damping material, and the outer layer made of insulating ceramic, which can effectively isolate external high-frequency noise and absorb reverberation sound, further reducing the interference of environmental noise on sound signals; The sound collection holes (2) are designed with a hexagonal honeycomb arrangement and a hole diameter between 1 mm and 3 mm, which can optimize the incident direction of the sound wave, reduce the interference between the holes, and improve the collection quality of the sound running inside the device; The sound collection hole (2) is a conical hole, which is gradually narrowed from the hole mouth to the hole end, and can enhance the directionality of the sound wave during the propagation process, and prevent the sound wave from being too divergent when transmitted to the focusing concave mirror (8), thereby affecting the focusing of the sound; The surface of the focusing concave mirror (8) is designed in a parabolic shape, with a reflectivity of more than 95%. Through the efficient sound reflection characteristics, the sound wave energy is concentrated on the sound collection module (6), thereby further increasing the strength of the sound signal and improving the quality of voiceprint feature extraction; The focusing concave mirror (8) is provided with a sound-absorbing layer on the back thereof, which can absorb the energy of the sound waves that are not reflected and reduce the influence of the redundant sound waves; The purpose is to achieve precise focusing on the sound of equipment operation through the array-type multi-directional sound collection module and the focusing concave mirror, reduce environmental noise interference, improve the quality of sound signals, and extract multi-dimensional voiceprint features to facilitate subsequent equipment operation status analysis.

[0015] It should be noted that, unless otherwise clearly specified and limited, the terms "placed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection method such as a folding connection, a rivet connection, a pin connection, a bonding connection and a welding connection, or a detachable connection method such as a threaded connection, a snap connection and a hinge connection, or an integral connection, or an electrical connection, or a direct connection, or an indirect connection through an intermediate medium, or the internal connection of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0016] It should be further pointed out that, when describing the above specific embodiment, for the sake of simplicity and clarity, only the differences between the above specific embodiment and other embodiments are described, but those skilled in the art should know that the above specific embodiment itself is also an independent technical solution.

Claims

1. A focused voiceprint collection sensor, characterized by: The invention is composed of a sound insulation cover, a sound collection hole, a voiceprint collection tube, a multi-angle voiceprint collection tube, a connecting rod, a sound collection module, a voice processing module and a focusing concave mirror. One end of the voiceprint collection tube is placed on the multi-angle voiceprint collection tube and is connected to the multi-angle voiceprint collection tube. The other end of the voiceprint collection tube is an open structure. The sound insulation cover is placed on the other end of the voiceprint collection tube. The sound insulation cover is provided with densely arranged sound collection holes. The sound collection holes are tiny holes. The sound collection holes are close to the upper edge of the sound insulation cover. The focusing concave mirror is placed in the voiceprint collection tube and corresponds to the sound collection holes. The sound collection module is placed in the voiceprint collection tube. The sound of the equipment operation is transmitted to the focusing concave mirror through the sound collection hole. The reflection focus is formed by the concave arc surface of the focusing concave mirror and falls on the sound collection module. The voice processing module is placed in the multi-angle voiceprint collection tube. The sound collection module transmits signals to the voice processing module through the signal line. The voice processing module is provided with a digital filter. After receiving the sound signal collected by the sound collection module, the voice processing module extracts voiceprint features.

2. The focused voiceprint acquisition sensor according to claim 1, characterized in that An insulating film is provided on the surface of the multi-angle voiceprint collection tube and the voiceprint collection tube, and the insulating film completely wraps the multi-angle voiceprint collection tube and the voiceprint collection tube.

3. The focused voiceprint acquisition sensor according to claim 1, characterized in that A collection sealing sleeve is arranged between the sound insulation cover and the voiceprint collection tube. The collection seal is made of a flexible material, which can reduce sound leakage, prevent external noise from entering the device from the joint, and improve the overall noise reduction effect.

4. The focused voiceprint acquisition sensor according to claim 1, characterized in that The voiceprint collection tube and the multi-angle voiceprint collection tube are detachably connected, and the surfaces of the voiceprint collection tube and the multi-angle voiceprint collection tube are coated with three layers of sound-absorbing layers, which absorb high-frequency, medium-frequency and low-frequency sounds respectively.

5. The focused voiceprint acquisition sensor according to claim 1, characterized in that A sound amplification module is provided on the inner wall of the sound insulation cover at a position corresponding to the sound collection hole, and the inner walls of the voiceprint collection tube and the multi-angle voiceprint collection tube are made of porous sound-absorbing material.

6. The focused voiceprint acquisition sensor according to claim 1, characterized in that There are multiple voiceprint collection tubes, and the multiple voiceprint collection tubes are arranged equidistantly along the circumference of the multi-angle voiceprint collection tube to form an array-type multi-directional sound collection, and two adjacent voiceprint collection tubes are not on the same horizontal plane.

7. The focused voiceprint acquisition sensor according to claim 1, characterized in that The sound insulation cover has a multi-layer composite structure, the innermost layer is made of high-density sound-absorbing cotton, the middle layer is a damping material, the outer layer is insulating ceramic, and the sound insulation cover is a concave structure.

8. The focused voiceprint acquisition sensor according to claim 1, characterized in that The sound collection hole adopts a hexagonal honeycomb arrangement design with an aperture between 1mm and 3mm. The sound collection hole is a conical hole that gradually narrows from the hole mouth to the hole end. The sound collection module is placed in the voiceprint collection tube through a connecting rod, and the connecting rod is a telescopic rod with adjustable length.

9. The focused voiceprint acquisition sensor according to claim 1, characterized in that The sound collection module is located between the focusing concave mirror and the sound collection hole, and the heights of the focusing concave mirror and the sound collection hole are higher than the sound collection module. The surface of the focusing concave mirror is designed in a parabolic shape with a reflectivity of more than 95%. A sound-absorbing layer is provided on the back of the focusing concave mirror.

10. The focused voiceprint acquisition sensor according to claim 1, characterized in that The voiceprint feature extraction method comprises the following steps: (1) Remove background noise and high-frequency noise through digital filters; (2) Normalize the amplitude of the sound signal to eliminate the impact of differences in sound intensity between different sensors or under different environmental conditions; (3) The continuous sound signal is divided into multiple time windows (i.e. frames), with the length of each frame ranging from 20 to 100 ms, and the Hanning window method is used to reduce inter-frame interference; (4) Perform fast Fourier transform (FFT) on each frame of sound signal to convert the time domain signal into a frequency domain signal; (5) Calculate the power spectral density (PSD) of the sound signal and draw a spectrum diagram to obtain the main frequency components and their energy distribution in the signal; (6) Extract the main frequency peaks and their corresponding amplitude and phase information from the spectrum diagram; (7) Calculate the mean and variance of the sound signal in each time window to characterize the fluctuation intensity of the signal; (8) Analyze the sharpness and asymmetry of the probability density function (PDF) of the sound signal and extract the kurtosis and skewness features; (9) The frequency component characteristics, energy distribution characteristics, fluctuation characteristics, kurtosis and skewness characteristics are integrated to construct a multi-dimensional voiceprint feature vector to comprehensively characterize the operating status of the device.

Images