CN104575500B - Application, speech recognition system and method for the electronic skin in speech recognition - Google Patents

Abstract

The invention discloses a voice recognition system, comprising: an electronic skin, the electronic skin includes a flexible sensitive layer, at least one of the upper surface and the lower surface of the sensitive layer is a non-planar structure, and the upper surface of the sensitive layer is An upper electrode layer and a lower electrode layer are respectively formed on the surface and the lower surface; the data processing module receives the signal from the electronic skin and sends the processed signal to the display module; the display module receives and displays the signal from the data processing module. The identification system of the present invention has the advantages of synchronous identification, high identification rate, small size and convenient portability. At the same time, due to the use of flexible film materials, the device can be easily integrated into a wearable electronic device.

Description

Application of electronic skin in speech recognition, speech recognition system and method

technical field

The application belongs to the field of sensors, and in particular relates to an application of electronic skin in speech recognition, a speech recognition system and a recognition method.

Background technique

With the rapid development of communication technology and the continuous advancement of computer science and technology, speech recognition is an eye-catching high-tech intelligent human-computer interaction technology, which involves phonetics, vocal mechanism, microelectronics technology, computer information processing technology, voice The application of multi-disciplinary comprehensive technologies such as signal processing technology, circuits and systems, and sensing technology has become a competitive emerging high-tech industry. As one of my country's strategic emerging industries, the intelligent speech recognition industry has always been highly valued by national authorities at all levels, and has been included in a number of national science and technology development plans and policy support areas.

Most of the currently reported speech recognition technologies are based on speech templates, large vocabulary continuous speech recognition, and acoustic models. However, these traditional speech recognition technologies face many problems. For example, in noisy environments, unclear pronunciations such as accents or dialects, and multiple people speaking at the same time, the voice input effect is poor, the recognition rate is not high, or even cannot be recognized. The main reason for the above problems is that the traditional voice acquisition module collects voice by capturing the propagation signal of voice in the air, and the interference of other external sound sources will affect the collected voice data.

Contents of the invention

The object of the present invention is to provide an application of electronic skin in speech recognition and a speech recognition system and method, so as to solve the problems of poor speech input effect and low recognition rate in the prior art.

To achieve the above object, the present invention provides the following technical solutions:

The embodiment of the present application discloses the application of electronic skin in speech recognition. The electronic skin includes a flexible sensitive layer, at least one of the upper surface and the lower surface of the sensitive layer is a non-planar structure, and the upper surface of the sensitive layer An upper electrode layer and a lower electrode layer are respectively formed on the surface and the lower surface.

The invention also discloses a speech recognition system, comprising:

Electronic skin, the electronic skin includes a flexible sensitive layer, at least one of the upper surface and the lower surface of the sensitive layer is a non-planar structure, the upper surface and the lower surface of the sensitive layer are respectively formed with an upper electrode layer and a lower electrode layer;

The data processing module receives the signal from the electronic skin and sends the processed signal to the display module;

The display module receives and displays the signal from the data processing module.

Preferably, in the above speech recognition system, the material of the sensitive layer is polydimethylsiloxane (PDMS), and the thickness ranges from 2 to 50 μm, preferably 50 μm.

Preferably, in the above speech recognition system, the material of the sensitive layer is a polymer material, and the polymer material is selected from polyethylene phthalate, polyvinyl alcohol, polyvinyl formal, polyethylene one or a combination of more.

Preferably, in the above speech recognition system, a flexible supporting layer is further arranged between the sensitive layer and the lower electrode layer.

Preferably, in the above speech recognition system, the support layer is a polyethylene (PE) film with a thickness in the range of 1-100 μm, preferably 12 μm.

Preferably, in the above speech recognition system, the material of the support layer can also be selected from highly transparent and flexible polyvinyl chloride (PVC) film, polyvinylidene chloride (PVDC), polyethylene terephthalate One or more combinations of alcohol ester (PET) films.

Preferably, in the above speech recognition system, the material of the upper electrode layer and the lower electrode layer is selected from gold, platinum, nickel, silver, indium, copper, aluminum, carbon nanotubes, graphene, and silver nanowires. One or more combinations.

Preferably, in the above-mentioned electronic skin, the electrodes of the upper electrode layer and the lower electrode layer are formed by evaporation, chemical deposition, printing, spraying or sputtering.

Preferably, in the above speech recognition system, the non-planar structure is a plurality of protruding polygonal pyramids.

Preferably, in the above speech recognition system, the bottom surface of each vertebral body is a square of 10 μm×10 μm, the angle between the side surface and the bottom surface is 54.7°, the height of the vertebral bodies is 7.06 μm, and the distance between the vertebral bodies is 1-100 μm, the distance between vertebral bodies is preferably 10 μm.

The invention also discloses a speech recognition method, comprising;

Using the above-mentioned electronic skin to collect vibration signals of the vocal cords;

Filter and amplify the collected vibration signal, and extract the characteristic signal;

The characteristic signal is analyzed and identified by a time-domain analysis method or a frequency-domain analysis method, and the identification result is displayed through a display module.

Compared with the prior art, the present invention has the advantages of:

The electronic skin of the present invention is integrated with the new micro-nano sensing technology. Since it uses a non-toxic and biocompatible ultra-thin elastic film material, it can be well integrated with human skin and constructed into a wearable skin. The type device can bring a super good user experience; at the same time, due to the unique nanostructure of the sensitive material, the device has higher sensitivity and better stability; in addition, the whole device is light, small and smart, and has the advantage of being easy to carry.

The speech recognition system obtained by the electronic skin of the present invention has the advantages of synchronous recognition, high recognition rate, small size, convenient portability and wearable.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments described in this application. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 shows the schematic diagram of speech recognition system in the specific embodiment of the present invention;

Figure 2 is a cross-sectional view of the electronic skin in a specific embodiment of the present invention;

Fig. 3 shows the SEM figure of the silicon wafer template in the specific embodiment of the present invention;

Figure 4 shows the SEM image of the PDMS flexible film made from the template shown in Figure 3;

FIG. 5 is a schematic diagram of speech recognition by the speech recognition system in the embodiment of the present invention.

Detailed ways

The current international competition in the information industry has increasingly manifested itself as a competition for science and technology. As a strategic and forward-looking important emerging technology industry, the speech recognition industry has always been the focus of attention of the scientific and industrial circles at home and abroad. The present invention relates to a new type of speech recognition technology and corresponding devices. The present invention proposes to use a micro-nano sensor to collect the vibration signal generated by the vocal cords during pronunciation, instead of capturing the propagation signal of speech in the air through the acquisition module as in the traditional speech recognition technology. Then the vibration signal is amplified, filtered and preprocessed, the characteristic signal is extracted and identified and analyzed by time domain analysis method, frequency domain analysis method, etc., so as to obtain a simple, feasible and systematic vibration signal identification method, and finally display the identification result.

The present invention innovates on the basis of traditional speech recognition technology, and mainly focuses on the collection method and extraction device of speech signals. It has the advantages of synchronous recognition, high recognition rate, small size and convenient portability. At the same time, due to the use of flexible film materials, The device can be easily integrated into a wearable electronic device.

The technical solutions in the embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

Referring to Fig. 1, in the embodiment of the present invention, the voice recognition system includes an electronic skin, a data processing module and a display module.

The electronic skin is used to collect the vibration signal of the vocal cords; the data processing module filters and amplifies the collected vibration signal, and extracts the characteristic signal, uses the time domain analysis method or the frequency domain analysis method to analyze and identify the characteristic signal, and passes the recognition result through Display module display.

2, the electronic skin in this embodiment includes a flexible support layer 1, a flexible sensitive layer 2 formed on the upper surface of the support layer 1, and a flexible sensitive layer 2 formed on the upper surface of the sensitive layer 2 and the lower surface of the support layer 1, respectively. Upper electrode layer 3 and lower electrode layer 4.

The material of the sensitive layer 2 is preferably PDMS (polydimethylsiloxane), and the material of the sensitive layer 2 can also be other polymer materials, such as polyethylene phthalate, polyvinyl alcohol, polyvinyl formal, Any one or more of polyethylene.

The upper surface of the sensitive layer 2 is a non-planar structure, preferably a plurality of raised vertebral bodies 21, the bottom surface of each vertebral body 21 is a square of 10 μm×10 μm, the angle between the side surface and the bottom surface is 54.7°, and the height of the vertebral body is is 7.06μm, and the distance between vertebral bodies is 10μm.

In other embodiments, the shape of the upper surface of the sensitive layer 2 may also be other non-planar structures such as waves.

It is easy to imagine that the lower surface of the sensitive layer 2 can also be set in a non-planar shape, and correspondingly, an electrode layer also needs to be set on the lower surface of the sensitive layer 2 .

The supporting layer 1 is preferably a highly transparent and highly flexible ultra-thin PE (polyethylene) film, and its thickness is preferably 12 μm.

The role of the PE film is to help the patterned PDMS film on the surface of the silicon wafer to be completely and easily separated from the template, and at the same time serve as a substrate to support the PDMS film.

The materials of the upper electrode layer 3 and the lower electrode layer 4 are selected from one or more combinations of gold, platinum, nickel, silver, indium, copper, aluminum, carbon nanotubes, graphene, and silver nanowires.

In the above-mentioned electronic skin, the overall thickness of the support layer 1 and the sensitive layer 2 is less than 70 μm. This ultra-thin electronic skin has the same softness and Young’s modulus as human skin, and is also ultra-light in weight. It can be directly bonded to the skin. A wearable device.

The manufacturing method of the above-mentioned electronic skin is as follows:

S1. Making templates that can form non-planar structures

In this embodiment, a template with a microstructure (preferably a 4-inch silicon wafer) is produced by using photolithography, etching and other processes in the MEMS manufacturing technology. Clean the surface of the silicon wafer first, then spin-coat photoresist, pre-baking, photolithography, development, and post-baking, and finally transfer the pattern to the silicon wafer by etching to obtain a template with a microstructure. The specific preparation method as follows:

1. Board making

After passing the design demonstration, use the drawing software to draw the drawings. Simply increasing the width and height of the pattern and the spacing between the patterns are conducive to improving the sensitivity of the device. However, after synthesis, the micropattern is preferably a pyramid structure, and the bottom surface of the tower is 10μm×10μm The angle between the side and the bottom is 54.7°, the distance between the top and the bottom of the tower is 7.06 μm, and the pattern pitch is 10 μm. Make a mask according to the size of the drawing.

2. Prepare a 4-inch silicon wafer

The silicon wafer is polished on one side of 4 inches, 300nm thermally oxidized SiO _2- layer silicon wafer on one side, MOS grade acetone and deionized water are ultrasonicated for 15 minutes, then MOS grade ethanol is ultrasonicated for 10 minutes and dried, and then baked at 105 °C Let dry for 10 minutes.

3. Patterned photoresist

a. Spin glue: Spin-coat 6-7um photoresist on the surface of the prepared 4-inch silicon wafer, preferably AZ4620, pre-rotate at 500rpm for 6s, and spin-coat at 400rpm for 30s;

b. Pre-baking: pre-baking at 95° for 210 seconds;

c. Exposure: Expose for 24 s in low vacuum mode using MA6 contact lithography machine;

d. Development: The ratio of the developer used is tetramethylammonium hydroxide: deionized water = 1:8, and the development time is 95 s;

e. Post-baking: post-baking at 95° for 180 seconds.

4. Graphic transfer to silicon wafer

a. Degumming: The plasma degumming machine removes the remaining photoresist film after development;

b. Dry etching: use reactive ion etching (RIE) to remove the patterned _SiO2 layer for 6 minutes

c. Wet directional etching: use 30% KOH solution to etch at 78°C for 9 minutes, and finally form an inverted quadrangular pyramid structure on the surface of the silicon wafer, as shown in Figure 3.

, forming a layer of organic molecules on the surface of the template

A thin layer of organic (trimethylchlorosilane or perfluorooctyltrichlorosilane) molecular layer is processed (such as vapor deposition or fumigation) on the surface of the template to ensure the integrity and ease of the PDMS film on the surface of the silicon wafer and the template. separate.

, Form a sensitive layer on the organic molecular layer

Then spin-coat (preferably at 3000r/min, time 30S) a transparent liquid polymer (such as polydimethylsiloxane) on the organic molecular layer, preferably the ratio of initiator and reactant is mass Ratio 1:10,) to make it evenly form a thin film (preferably 50μm in thickness).

, forming a support layer on the sensitive layer

Then, a layer of highly transparent and flexible ultra-thin PE (polyethylene) film (thickness is preferably 12 μm) is formed on the surface of the above film seamlessly without bubbles (bubbles and gaps are also included).

, heat treatment, peel off the cured sensitive layer and support layer from the template

Then heat it in a vacuum environment (preferably at 65-75°C) for a period of time (preferably 2-3 hours) and wait until the above-mentioned liquid polymer PDMS film is completely cured, and at the same time it is completely integrated with the PE film. The cured polymer film is peeled off from the surface of the silicon wafer template, so that the micropattern on the silicon wafer template is replicated on the PDMS flexible film to prepare a film with a pyramidal microstructure. The overall thickness of the film is <70 μm, such as Figure 4 shows.

, forming an upper electrode layer and a lower electrode layer on the upper surface of the sensitive layer and the lower surface of the supporting layer respectively

Finally, uniform coating (such as evaporation, chemical deposition, etc., this patent preferably evaporates) a layer of ultra-thin nano-conductive film (Au particles with a purity of 99.9999% in this invention, the thickness of the evaporated Au conductive film is uniformly coated on the upper and lower surfaces of the film. 100nm), at this time, an ultra-thin flexible conductive electronic skin with multiple sensitive sites is formed.

Then a flexible electrode (such as enameled wire with a diameter of 0.1mm, 20μm thick and with pressure-sensitive adhesive) is drawn out (such as bonding, welding, etc., this patent is preferably bonding) on the Au nano-conductive layer on the upper and lower layers of the ultra-thin flexible conductive electronic skin. flat copper foil tape, flexible and ultra-thin (10μm thick) copper foil, this patent is preferably flexible and ultra-thin copper foil).

Finally, a layer of PDMS is spin-coated (preferably at 5000r/min, time 30S) to uniformly form a thin (thickness <10μm) protective layer of Au conductive layer, and a capacitive ultra-thin flexible electronic skin is constructed. .

Select a piece of ultra-thin flexible electronic skin, physically cut out 2 small pieces of different sizes (preferably 2 pieces of 15mm×15 mm and 20mm×15 mm), and then fix the flexible wires on one edge of the patterned surface of these small pieces respectively To make electrodes, and finally on a transparent substrate (such as glass, PDMS film, etc., this patent prefers a PDMS film with a thickness of 2mm) to seamlessly bond the two patterned surfaces to form an ultra-thin wearable flexible device .

When the device is on the bare chip, use a resistance measuring tool to measure its basic impedance, and at the same time add a small voltage (preferably 2-5V); use an oscilloscope to test its background noise when the device is in an unloaded state, and then collect the vibration signal of the vocal cords during pronunciation It is found that the background noise is too large, and the vibration signal cannot be effectively screened out from the noise. After testing and analyzing the external circuit and optimizing the device, when the device is placed on the vocal cord, the impedance of the sensor will change greatly due to the vibration of the vocal cord, and the range of voltage fluctuations generated on the sampling resistor is higher than the background measured by the oscilloscope Noise, the resulting voltage transition level can be separated from the noise level and identified by the test circuit. Then use the filtering device in the circuit to filter out the background noise signal to obtain an effective pulse wave signal. After the signal is amplified by the amplifier circuit, the signal is transmitted to the receiving device on the display using wireless Bluetooth transmission technology to synchronize the vocal fold vibration wave during pronunciation. displayed on the display.

As shown in Figure 5, paste the device on the vocal cords. When we speak and pronounce, the vocal cords will vibrate accordingly, and the sensor will automatically collect the vibration signals. After the conversion and recognition of the external circuit, it will be transmitted to the display using Bluetooth (preferably 7-inch pad ), the waveform on the display will show waveforms with different amplitudes and frequencies according to the severity and length of the pronunciation.

It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is a relationship between these entities or operations. There is no such actual relationship or order between them. Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, but also includes elements not expressly listed. other elements of or also include elements inherent in such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

The above description is only the specific implementation of the present application. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present application, some improvements and modifications can also be made. It should be regarded as the protection scope of this application.

Claims (12)

1. application of the electronic skin in speech recognition, the electronic skin includes sensitive layer flexible, the sensitive layer At least one upper and lower surface is nonplanar structure, is respectively coated and is formed in the upper and lower surface of the sensitive layer Upper electrode layer and lower electrode layer, wherein the upper electrode layer and/or lower electrode layer formed on the nonplanar structure is also non-flat Face structure.

2. a kind of speech recognition system, which is characterized in that including：

Electronic skin, the electronic skin include sensitive layer flexible, the upper and lower surface of the sensitive layer at least it One is nonplanar structure, is respectively coated in the upper and lower surface of the sensitive layer and is formed with upper electrode layer and lower electrode layer, Wherein, the upper electrode layer and/or lower electrode layer formed on the nonplanar structure is also nonplanar structure；

Data processing module, receives the signal from electronic skin and signal is sent to display module by treated；

Display module receives and shows the signal from data processing module.

3. speech recognition system according to claim 2, it is characterised in that：The material of the sensitive layer is poly dimethyl silicon Oxygen alkane, thickness range are 2~50 μm.

4. speech recognition system according to claim 2, it is characterised in that：The material of the sensitive layer is macromolecule material Material, the high molecular material in poly- polyethylene terephthalate, polyvinyl alcohol, polyvinyl formal, polyethylene one Kind or a variety of combinations.

5. speech recognition system according to claim 2, it is characterised in that：Between the sensitive layer and lower electrode layer also It is provided with supporting layer flexible.

6. speech recognition system according to claim 5, it is characterised in that：The supporting layer is polyethylene film, thick Ranging from 1~100 μm of degree.

7. speech recognition system according to claim 5, it is characterised in that：The support layer material is also selected from height Polyvinyl chloride film, the polyvinylidene chloride of transparent high flexibility, it is one or more in pet film Combination.

8. speech recognition system according to claim 2, it is characterised in that：The upper electrode layer and lower electrode layer material One or more combinations in gold, platinum, nickel, silver, indium, copper, aluminium, carbon nanotube, graphene, nano silver wire.

9. speech recognition system according to claim 2, it is characterised in that：The upper electrode layer and lower electrode layer electrode It is to be formed by way of vapor deposition, chemical deposition, printing, spraying or sputtering.

10. speech recognition system according to claim 2, it is characterised in that：The nonplanar structure be protrude out it is more A more pyramid bodies.

11. speech recognition system according to claim 10, it is characterised in that：The bottom surface of each centrum is 10 μm The angle of × 10 μm of square, side and bottom surface is 54.7 °, and vertebral height is 7.06 μm, and spacing is 1~100 μ between centrum m。

12. a kind of audio recognition method, which is characterized in that including；

The vibration signal of vocal cords is acquired using any electronic skin of claim 2 to 11；

The vibration signal of acquisition is filtered, is amplified, and extracts characteristic signal；

Analysis and identification is carried out to characteristic signal using temporal analysis or frequency domain analysis, and recognition result is passed through into display module Display.