Optical artificial laryngeal device and application method thereof

Abstract

The present invention discloses an optical artificial laryngeal device and its application method, comprising: a signal providing unit for providing the device with an original optical path signal; a sensing unit for sensing and acquiring the user's throat sound vibration information and head movement information; a sensitivity enhancing unit for assisting in enhancing the sensitivity of the information acquired by the sensing unit; a signal receiving unit for receiving an optical path signal containing the user's throat sound vibration information and head movement information after passing through the sensing unit and the sensitivity enhancing unit, which has undergone optical power loss, and performing signal demodulation to convert the optical path signal into intensity information; and a data processing unit for processing and analyzing the intensity information to acquire the user's semantic expression information and represent it on a computer to achieve silent speech recognition. While achieving improvements in comfort, durability, and recognition accuracy, the present invention also provides a more diversified expression method for special groups of patients.

Description

Technical Field

[0001] The present invention belongs to the technical field of voice assistance and human-computer interaction, and in particular relates to an optical artificial laryngeal device and an application method thereof. Background Art

[0002] Language communication is one of the most important means of communication in human society. People use language to communicate, obtain information, and cope with daily life. However, some special groups, such as patients with laryngeal diseases, vocal cord dysfunction, and laryngeal cancer surgery, have lost their normal ability to speak due to congenital diseases, trauma, or surgery. They can only express their ideas through other means, such as through sign language or writing. However, sign language is not suitable for communicating with the general public, and writing is too inefficient and inconvenient for some young people or those with low cultural levels. In general, this loss of language function has a huge impact on patients' daily life, work, and social interactions, seriously reducing their quality of life and mental health.

[0003] With the development of human-computer interaction technology, researchers are paying more and more attention to this group and developing related technologies to help them communicate. For example, sign language translation gloves, which are sensor-based, use multi-channel stress and strain sensors to recognize multiple gestures, helping people with special needs to communicate with the general public through sign language. However, multiple sensors also mean that the system is highly dependent on each sensor, which is inconvenient and durable. If a sensor fails, the entire system will fail. There are also cameras that recognize facial expressions and mouth shapes, achieving true "lip reading", but this system also has the shortcomings of other camera systems. If the body moves, the recognition accuracy will decrease, or even fail to recognize.

[0004] Other researchers have focused on more direct ways of expression, using throat vibrations to express meaning. This method does not require the deaf and mute to actually make a sound, but only requires throat vibrations to express, which is more direct and less likely to cause signal loss. For example, recognition can be achieved through throat vibration microphone headphones, or multimodal recognition can be achieved by combining microphones, speakers, word processors, and sound processors. However, these results are not very accurate due to the limitations of the equipment themselves. Surface electromyography, as a highly sensitive recognition technology, can infer the user's intentions by capturing oral, throat, or neck muscle movement signals. However, it is also greatly affected by environmental noise, and the signal processing is complex, which places high demands on device stability and wearing comfort.

[0005] To address the above technical issues, the new technology should have high sensitivity to obtain and process information such as weak throat vibrations; high stability to be suitable for complex scenarios (electromagnetic interference, etc.); high biocompatibility to prevent allergic reactions; and finally, it should be comfortable and durable, with high comfort and long-term use. Summary of the Invention

[0006] The purpose of the present invention is to propose an optical artificial laryngeal device and semantic recognition method for vocal assistance. By integrating a specific structure fiber optic sensor with a sensitivity enhancement unit, vibration information at the throat and head movement information are collected, and through experimental data training combined with a neural network, high-accuracy speech signal recognition can be achieved without making any sound, helping special groups to express their ideas without obstacles.

[0007] To achieve the above-mentioned objectives, the present invention provides an optical artificial larynx device and an application method thereof. An optical artificial larynx device comprises:

[0008] A signal providing unit, used to provide the original optical path signal to the device;

[0009] A sensing unit, connected to the signal providing unit, for sensing and obtaining throat vocal vibration information and head movement information of the user; wherein the sensing unit comprises a fiber optic sensor with a specific structure and a special outer coating material, Dragon Skin 20; the fiber optic sensor with a specific structure is a liquid-core polymer composite fiber;

[0010] a sensitivity enhancement unit, connected to the sensing unit, for assisting in enhancing the sensitivity of information acquired by the sensing unit;

[0011] a signal receiving unit connected to the sensing unit, configured to receive an optical path signal containing throat vocal vibration information and head movement information of the user after passing through the sensing unit and the sensitization unit and experiencing optical power loss, and perform signal demodulation to convert the optical path signal into intensity information;

[0012] The data processing unit is connected to the signal receiving unit and is used to process and analyze the intensity information, obtain user semantic expression information, and represent it on a computer to achieve silent speech recognition.

[0013] Preferably, the liquid core polymer composite optical fiber comprises a liquid core polymer optical fiber and a polymethyl methacrylate polymer optical fiber;

[0014] Among them, the two outer parts are polymethyl methacrylate polymer optical fibers, the inner core is liquid core polymer optical fiber, and ultraviolet glue is used to seal the joints to form a complete liquid core polymer composite optical fiber.

[0015] Preferably, the cladding portion of the liquid-core polymer optical fiber is a silicone tube made of sulfur-cured silicone rubber, and glycerol is selected as the light transmission medium in the liquid core portion.

[0016] Preferably, the inner diameter of the liquid-core polymer optical fiber is 0.3 mm, the length is 20.0 mm, and the spacing between the two layers of polymethyl methacrylate polymer optical fiber is 2.0 mm.

[0017] Preferably, the sensitization unit comprises a sensitization ring and an elastic textile belt;

[0018] The sensitization ring is a circular ring structure with a cylinder running through the middle;

[0019] The cylinder is the core part of the sensitivity enhancement, which is used to bond with the sensing unit and apply stress to increase the sensitivity of the sensing unit, and is fixed on the elastic textile belt. Through the stretching of the elastic textile belt, the wearable device is achieved while further enhancing the sensitivity of signal perception.

[0020] Preferably, the sensitivity enhancement ring is made of SUS316 stainless steel;

[0021] The elastic textile belt is made of velvet material and has a hole reserved therein;

[0022] The sensitization ring is embedded in the hole to achieve the wearing around the neck and sensitization effects.

[0023] Preferably, the diameter of the circular ring structure of the sensitization ring is 3.0 cm, wherein the outer diameter is 2.0 mm and the inner diameter is 1.0 mm.

[0024] Preferably, the data processing unit includes a back-end computer and corresponding processing software and algorithms;

[0025] The back-end computer extracts the demodulated information of the signal receiving unit through the network cable, and further processes and analyzes the demodulated information, and performs multi-modal semantic analysis and expression through direct recognition and Morse code analysis.

[0026] Preferably, the algorithm comprises a neural network BKA-DNN algorithm.

[0027] The present invention also provides an application method of the optical artificial laryngeal device, comprising:

[0028] An optical path signal containing semantic information is obtained through a signal receiving unit and a data processing unit, and features are extracted from the optical path signal. The extracted features are input into the model as model parameters for further analysis; wherein the features include time domain features and frequency domain features; the time domain features include mean, variance, skewness, kurtosis, and root mean square; the frequency domain features include dominant frequency and spectral entropy;

[0029] Perform semantic recognition based on the pre-trained BKA-DNN neural network model to obtain recognition results;

[0030] Among them, the semantic recognition process includes two discrimination methods. One is to directly identify the throat vocalization characteristics and directly compare the similarity with the semantic information contained in the pre-trained BKA-DNN neural network model; the other is to identify the meaning of the user's head expression by recognizing head movements according to the Morse code encoding method.

[0031] Compared with the prior art, the present invention has the following advantages and technical effects:

[0032] The present invention provides a new type of artificial larynx device, which is made of optical fiber materials with higher comfort, strong anti-electromagnetic interference ability and high sensitivity. It is combined with a sensitization unit to achieve wearability and realize higher accuracy recognition of user expression meaning through stress changes.

[0033] The present invention also provides a new method for semantic recognition and expression based on an optical artificial larynx device. This method uses a neural network to identify both laryngeal vibrations and subtle head movements. Compared to traditional methods that rely solely on laryngeal vibrations or laryngeal electromyography, this method offers higher accuracy and improved system stability and redundancy, providing users with a more diverse range of ways to express themselves accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0035] Figure 1 Schematic diagram of the structure of an optical artificial laryngeal device according to an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of an optical artificial laryngeal device according to an embodiment of the present invention;

[0037] Figure 3 Schematic diagram of the structure of the sensing unit and the sensitization unit according to an embodiment of the present invention;

[0038] Figure 4 A schematic diagram of the preparation process of the sensing unit and the sensitization unit according to an embodiment of the present invention;

[0039] Figure 5 Schematic diagram of a flow chart of a new method for semantic recognition and expression based on an optical artificial laryngeal device according to an embodiment of the present invention;

[0040] Figure 6A schematic diagram of the visual representation of Mandarin tones and English letters in the direct expression model of an embodiment of the present invention;

[0041] Figure 7 Schematic diagram of visual representation of Chinese and English short sentences in the direct expression model of an embodiment of the present invention;

[0042] Figure 8 Schematic diagram of various head movement tests in the Morse code model according to an embodiment of the present invention;

[0043] Figure 9 Schematic diagram of visual representation of four signal instances in the Morse code model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0045] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0046] like Figure 1-2 As shown, this embodiment provides an optical artificial laryngeal device, comprising:

[0047] Signal providing unit, sensing unit, sensitization unit, signal receiving unit, data processing unit;

[0048] The signal providing unit is directly connected to one end of the sensing unit and is used to provide a light source, that is, to provide the original optical path signal for the device;

[0049] The sensor unit, composed of a fiber optic sensor with a specific structure and a special outer coating material called Dragon Skin 20, is used to sense the user's throat vibration information and subtle head movements;

[0050] A sensitivity enhancement unit, including a sensitivity enhancement ring and an elastic textile belt, is used to assist the sensing unit and enhance the sensitivity of the sensing unit in acquiring information, thereby improving recognition accuracy;

[0051] The signal receiving unit is directly connected to the other end of the sensing unit and is used to receive the optical signal that has lost optical power after passing through the sensing and sensitization units, that is, the optical path signal containing the user's vibration information and head movement information, and demodulate it into intensity information;

[0052] The data processing unit, which consists of a back-end computer and corresponding processing software and algorithms, is used to process and analyze the intensity information obtained by the signal receiving unit, obtain user semantic expression information, and represent it on the computer to achieve silent speech recognition.

[0053] Specifically, the signal providing unit and receiving unit are composed of a light source SLD-Msc-371-HP2-SM and a spectrometer (Thorlab), respectively. The light source SLD-Msc-371-HP2-SM has the characteristics of high stability, high efficiency and low noise, which also means high stability and low noise interference for the entire optical artificial larynx. It is used to provide optical signals and transmit the optical signals into the optical fiber. The other end of the optical fiber is connected to the spectrometer, which demodulates the optical signals and outputs them as waveform parameter information for further analysis.

[0054] Furthermore, the specific structure optical fiber sensor is a liquid core polymer composite optical fiber, which is composed of a liquid core polymer optical fiber and a polymethyl methacrylate polymer optical fiber. The two outer parts are polymethyl methacrylate polymer optical fibers, the inner core is a liquid core polymer optical fiber, and ultraviolet glue is used to seal the joints to form a complete liquid core polymer composite optical fiber; the cladding part of the liquid core polymer optical fiber is a silicone tube made of sulfur-cured silicone rubber, and the liquid core part selects glycerol with a high refractive index as the light transmission medium.

[0055] When the throat vibrates or the head nods or makes slight movements, the liquid-core polymer optical fiber at that location is subjected to stress and bends, resulting in optical power loss in the optical signal. That is, the movement of the throat and head causes changes in the corresponding signals. Conversely, the characterization information of the throat and head can also be obtained by analyzing the signal changes.

[0056] Optionally, the inner diameter of the liquid-core polymer optical fiber is 0.3 mm, the length is 20.0 mm, and the spacing between the two layers of polymethyl methacrylate polymer optical fiber is 2.0 mm.

[0057] The steps of manufacturing the liquid core polymer composite optical fiber include polishing the end face of the polymethyl methacrylate polymer optical fiber, preparing the liquid core polymer optical fiber, assembling the liquid core polymer composite, and outer coating and packaging.

[0058] Furthermore, the sensitization unit consists of a sensitization ring and an elastic textile belt; wherein, the sensitization ring is a circular ring structure with a cylinder running through the middle. The cylinder is the core part of the sensitization, which is used to bond with the sensing unit and apply stress to increase the sensitivity of the sensing unit, and is fixed on the elastic textile belt. By stretching the elastic textile belt, the wearable device is achieved while further enhancing the sensitivity of signal perception.

[0059] Furthermore, the sensitivity enhancement ring is made of SUS316 stainless steel;

[0060] The elastic textile belt is made of velvet material and has a reserved hole. The sensitivity-enhancing ring is embedded in the hole, and the combination of the two can achieve the effects of wearing around the neck and enhancing sensitivity.

[0061] The diameter of the ring structure of the sensitization ring is 3.0 cm, the outer diameter is 2.0 mm, and the inner diameter is 1.0 mm.

[0062] Furthermore, the data processing unit is composed of a back-end computer and corresponding processing software algorithms. The computer extracts the demodulated information of the signal receiving unit through the network cable, and further processes and analyzes the information, realizing multi-modal semantic analysis and expression through direct recognition and Morse code analysis.

[0063] Optionally, the algorithm of the data processing unit mainly uses the neural network BKA-DNN algorithm.

[0064] like Figure 3 The figure shows a detailed schematic diagram of the sensor unit and the sensitization unit, the main components of the optical artificial laryngeal device. The core of the sensing unit is a liquid-core polymer composite fiber, composed of a liquid-core polymer fiber and a polymethyl methacrylate polymer fiber. Laryngeal vibration and head movement generate stress changes that cause bending in the liquid-core polymer composite fiber, leading to optical power loss and corresponding signal changes.

[0065] The outer layer's covering material, Dragon Skin 20, is a special material with an elastic modulus close to that of human tissue. It can not only ensure good fit with the human body and comfort, but also prevent changes in the optical fiber combination structure, ensuring the durability of the optical fiber combination structure and avoiding external environmental pollution.

[0066] The outer covering material and the sensitizing ring of the sensitizing unit are glued together to ensure the stability of the connection.

[0067] The sensitivity enhancement ring is made of SUS316 stainless steel, which has high strength and corrosion resistance. The ring has a diameter of 3.0 cm, an outer diameter of 2.0 mm, and an inner diameter of 1.0 mm. It is designed as a protruding structure and is glued to the covering material of the optical fiber sensor using Baide glue to improve the pressure perception sensitivity of the optical fiber by stress concentration. Under conditions of 0 to 50 mN, the pressure sensitivity of the liquid-core polymer composite optical fiber without an integrated sensitivity enhancement ring is only 0.34 Mw / mN, while the pressure sensitivity of the optical fiber with an integrated sensitivity enhancement ring reaches 1.92 Mw / mN.

[0068] The elastic textile belt is made of velvet material, which is soft, lightweight, wear-resistant and highly elastic. It can fit the skin of the neck well and will not feel burden or discomfort even after long-term use. A hole is pre-opened in the textile belt to adapt to the circular structure of the sensitization ring and be embedded in it, enhancing structural stability, making it wearable while further enhancing the sensitivity of the sensor unit.

[0069] like Figure 4 As shown, the main steps for the preparation and combination of the sensing unit and the sensitizing unit are provided.

[0070] The preparation process of liquid core polymer composite optical fiber includes:

[0071] Use optical fiber polishing paper (0.5 μm particle size) to polish the end face of the polymethyl methacrylate polymer optical fiber to ensure the flatness and smoothness of the optical fiber interface for subsequent connection;

[0072] Use a syringe to inject high-refractive-index glycerol into the sulfur-cured silicone rubber tube to form a liquid-core polymer optical fiber. Quickly connect the polished polymethyl methacrylate polymer optical fiber to the two ends of the liquid-core polymer optical fiber to ensure a tight connection. Apply ultraviolet glue (UV glue) to the interface and use ultraviolet curing equipment to cure the interface to form a complete sealing structure.

[0073] For the outer covering of the optical fiber, a circular mold with a diameter of 3.0 cm was made through 3D printing technology and printed with PLA material. The completed liquid core polymer optical fiber and polymethyl methacrylate polymer optical fiber structure were fixed in the center of the mold. Dragon Skin 20 elastomer material was mixed in a 1:1 ratio and poured into the mold to cover the entire optical fiber structure. Finally, it was heated and cured at 50°C for 2 hours. After cooling to room temperature, it was demolded to complete the outer covering of the optical fiber.

[0074] The production process of the sensitization ring and textile belt includes:

[0075] The sensitivity ring is made of SUS 316 stainless steel and manufactured by 3D printing. It is designed with an outer ring with an inner cylindrical protrusion structure with a diameter of 3cm, an outer diameter of 2.0mm, and an inner diameter of 1.0mm to ensure its corrosion resistance and high strength.

[0076] The textile belt is made of wear-resistant and lightweight velvet material to ensure wearing comfort and firmness, and holes are opened on the textile belt to facilitate combination with the sensitivity ring;

[0077] The unit synthesis and packaging process includes:

[0078] Insert both sides of the sensitizing ring into the holes reserved in the elastic textile belt, apply Baide glue on the central cylindrical surface of the sensitizing ring, and firmly fix the coated and encapsulated liquid core polymer composite optical fiber unit to the sensitizing ring to ensure a close combination of the sensitizing ring and the optical fiber;

[0079] Furthermore, the package integrity is checked to ensure that the liquid core inside the light has not leaked. After the inspection is correct, the two ends of the optical fiber are fixed to the textile tape in a cross-stitching manner using needle and thread to reduce the impact of accidental contact between the two ends of the optical fiber and the body on performance and ensure the stability of the system.

[0080] like Figure 5 As shown, based on the same inventive concept, this embodiment also provides an application method of an optical artificial laryngeal device, comprising:

[0081] The optical path signal containing semantic information is acquired by the signal receiving unit and the data processing unit, and is directly demodulated by the spectrometer of the signal receiving unit and converted into intensity parameter information, which is transmitted to the data processing unit, from which effective information is extracted for further processing and analysis. The optical path signal containing semantic information is the optical path signal containing relevant laryngeal vibration and head movement information;

[0082] The optical signal is received from the signal receiving unit and processed by the data processing unit. Five time domain features (mean, variance, skewness, kurtosis and root mean square) and two frequency domain features (main frequency and spectral entropy) are extracted from the signal receiving unit as model parameters and input into the model for the next step of analysis.

[0083] Perform semantic recognition based on the pre-trained BKA-DNN neural network model to obtain recognition results;

[0084] Specifically, the BKA-DNN algorithm is an efficient meta-heuristic optimization algorithm that enhances the algorithm's global search capability and ability to avoid local optimality by simulating the migration and predation behavior of black-winged kites and combining it with an innovative Cauchy mutation strategy. In the BKA-DNN algorithm, the full-time and bias of the deep neural network (DNN) are dynamically predicted and updated by BKA to minimize prediction errors and optimize algorithm performance. In addition, the algorithm also cleverly integrates a leadership strategy, using the current optimal solution information to guide the search direction, improve search efficiency, effectively balance global and local searches, and avoid premature convergence. This capability also makes the BKA-DNN algorithm extremely suitable for the target recognition and matching tasks of this model.

[0085] Furthermore, the pre-trained BKA-DNN neural network model includes two discrimination methods:

[0086] One method is to directly identify the characteristic information of laryngeal vocalization and directly compare the similarity with the semantic information contained in the pre-trained BKA-DNN neural network model;

[0087] The similarity comparison process is specifically carried out by performing global and local searches on the pre-trained model library to match the current information with the existing information of the model. After the matching is completed, the corresponding semantic features are extracted and output;

[0088] The other is to recognize head movements. This method is based on Morse code and does not require global or local searches of large amounts of information. Instead, it only needs to recognize head movements to match short dot signals and long signals in Morse code, and then compare the recognized dot signal and long signal combination with the Morse code table to identify the user's expression. Compared with the former, although it is slightly more difficult to use, the simplified model will make the recognition accuracy higher.

[0089] In order to realize model training, in this embodiment, 6 special population users and 3 general population users were recruited for data set training, and each training example had 100 sample signals, 80 of which were used as training sets, and the remaining 20 sample signals were used as test sets for model testing.

[0090] First, the model was trained and tested on the four tones of Mandarin and the five vowels A, E, I, O, and U in English, i.e. a total of nine training examples and 8,100 sample signals, of which 1,620 sample signals were used for testing.

[0091] refer to Figure 6 , provides recognition visualization representations of 9 training examples. It can be seen from the diagram that all 9 training examples have their own unique characteristics, and the recognition accuracy of the test set is as high as 97.21%, which meets the performance requirements.

[0092] Furthermore, a more complex model training was conducted using the same dataset with the same six users from a specific demographic. This training included six training examples: "Feel," "Good," "Feel Good," "Happy New Year," "Feel Good," and "Happy New Year." These examples included English words, Chinese phrases, and English phrases, effectively testing the model's ability to recognize complex semantics.

[0093] refer to Figure 7 , provides recognition visualization representations of 6 training examples. It can be seen from the diagram that all 6 training examples have unique features, and the recognition accuracy rate has reached 94.43%. Combined with the above-mentioned 97.21% accurate recognition rate of letter tones, it proves the effectiveness of the direct recognition model of laryngeal features.

[0094] The Morse code recognition model is also trained in a similar way. First, six users from special groups and three users from general groups are required to repeat the above actions twice in the order of screaming, nodding upwards, nodding downwards, swallowing, humming, and coughing to verify the model's action recognition ability.

[0095] refer to Figure 8 , providing a visual representation of model recognition. Six actions were clearly demonstrated, demonstrating the artificial larynx's ability to monitor minute movements, as well as its high stability and accuracy, which also lays the foundation for further Morse code encoding applications.

[0096] refer to Figure 9 , provides the actual recognition effects of four types of Morse code, requiring users to use two actions, small nods and large nods, to represent the four signals SOS, SAVE, DOWN and HELP according to a specific coding method. The visual legend also shows that the four signals have unique characteristics, and the recognition accuracy rate is as high as 99.39%, verifying the effectiveness of the Morse code recognition model.

[0097] The fiber optic sensor used in this embodiment has high flexibility and bendability due to its specific structure, and is very suitable as a monitoring element for the neck. While ensuring high sensitivity, it can be used for a long time and is not easy to break or damage. At the same time, fiber optic sensing naturally has anti-electromagnetic interference capabilities, biocompatibility and good chemical stability, which also means higher system stability than monitoring methods such as electromyography and skin electricity. In addition, the special material coating the optical fiber also makes the artificial throat device skin-friendly, can directly contact the skin and is also comfortable, suitable for helping users use it for a long time every day.

[0098] The sensitivity enhancement structure adopted in this embodiment also innovatively improves the perception sensitivity of the optical fiber sensor. The simple structure does not affect other performance of the system. At the same time, it takes into account both comfort and sensitivity while achieving wearability.

[0099] The new method for semantic recognition and expression based on an optical artificial larynx device proposed in this embodiment is essentially a customized recognition method based on the ability of the optical artificial larynx to recognize two modal information: throat vibration and head movement. Its innovation lies in the realization of multi-type information recognition through one device, and the realization of two diversified analysis methods: direct recognition and Morse code encoding recognition. At the same time, it also provides users with diversified expression methods, which is not only suitable for a wider range of people, but also takes into account the needs of more special groups, and improves the stability and redundancy of the system. When a single method cannot express the meaning, another method can be used as a good supplement. It has important practical significance for special groups of users to express their own voices.

[0100] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. An optical artificial laryngeal device, characterized in that: include: A signal providing unit, used to provide the original optical path signal to the device; A sensing unit, connected to the signal providing unit, for sensing and obtaining throat vocal vibration information and head movement information of the user; wherein the sensing unit comprises a fiber optic sensor with a specific structure and a special outer coating material DragonSkin 20; the fiber optic sensor with a specific structure is a liquid-core polymer composite fiber; a sensitivity enhancement unit, connected to the sensing unit, for assisting in enhancing the sensitivity of information acquired by the sensing unit; a signal receiving unit connected to the sensing unit, configured to receive an optical path signal containing throat vocal vibration information and head movement information of the user after passing through the sensing unit and the sensitization unit and experiencing optical power loss, and perform signal demodulation to convert the optical path signal into intensity information; The data processing unit is connected to the signal receiving unit and is used to process and analyze the intensity information, obtain user semantic expression information, and represent it on a computer to achieve silent speech recognition.

2. The optical artificial laryngeal device according to claim 1, characterized in that The liquid core polymer composite optical fiber comprises a liquid core polymer optical fiber and a polymethyl methacrylate polymer optical fiber; Among them, the two outer parts are polymethyl methacrylate polymer optical fibers, the inner core is liquid core polymer optical fiber, and ultraviolet glue is used to seal the joints to form a complete liquid core polymer composite optical fiber.

3. The optical artificial laryngeal device according to claim 2, characterized in that The cladding portion of the liquid core polymer optical fiber is a silicone tube made of sulfur-cured silicone rubber, and glycerol is selected as the light transmission medium in the liquid core portion.

4. The optical artificial laryngeal device according to claim 2, characterized in that The inner diameter of the liquid core polymer optical fiber is 0.3 mm, the length is 20.0 mm, and the spacing between the two layers of polymethyl methacrylate polymer optical fiber is 2.0 mm.

5. The optical artificial laryngeal device according to claim 1, characterized in that The sensitization unit includes a sensitization ring and an elastic textile belt; The sensitization ring is a circular ring structure with a cylinder running through the middle; The cylinder is the core part of the sensitivity enhancement, which is used to bond with the sensing unit and apply stress to increase the sensitivity of the sensing unit, and is fixed on the elastic textile belt. Through the stretching of the elastic textile belt, the wearable device is achieved while further enhancing the sensitivity of signal perception.

6. The optical artificial laryngeal device according to claim 5, characterized in that The sensitivity enhancement ring is made of SUS316 stainless steel; The elastic textile belt is made of velvet material and has a hole reserved therein; The sensitization ring is embedded in the hole to achieve the wearing around the neck and sensitization effects.

7. The optical artificial laryngeal device according to claim 5, characterized in that The diameter of the circular ring structure of the sensitization ring is 3.0 cm, wherein the outer diameter is 2.0 mm and the inner diameter is 1.0 mm.

8. The optical artificial laryngeal device according to claim 1, characterized in that The data processing unit includes a back-end computer and corresponding processing software and algorithms; The back-end computer extracts the demodulated information of the signal receiving unit through the network cable, and further processes and analyzes the demodulated information, and performs multi-modal semantic analysis and expression through direct recognition and Morse code analysis.

9. The optical artificial laryngeal device according to claim 8, characterized in that The algorithm includes the neural network BKA-DNN algorithm.

10. An application method of the optical artificial laryngeal device according to any one of claims 1 to 9, characterized in that: include: An optical path signal containing semantic information is obtained through a signal receiving unit and a data processing unit, and features are extracted from the optical path signal. The extracted features are input into the model as model parameters for further analysis; wherein the features include time domain features and frequency domain features; the time domain features include mean, variance, skewness, kurtosis, and root mean square; the frequency domain features include dominant frequency and spectral entropy; Perform semantic recognition based on the pre-trained BKA-DNN neural network model to obtain recognition results; Among them, the semantic recognition process includes two discrimination methods. One is to directly identify the throat vocalization characteristics and directly compare the similarity with the semantic information contained in the pre-trained BKA-DNN neural network model; the other is to identify the meaning of the user's head expression by recognizing head movements according to the Morse code encoding method.

Images