Abstract: A device includes first and second gate structures respectively extending across the first and second fins, and a gate isolation plug between a longitudinal end of the first gate structure and a longitudinal end of the second gate structure. The gate isolation plug comprises a first dielectric layer and a second dielectric layer over the first dielectric layer. The first dielectric layer has an upper portion and a lower portion below the upper portion. The upper portion has a thickness smaller than a thickness of the lower portion of the first dielectric layer.

Abstract: The present invention provides a method for OSA (Obstructive Sleep Apnea) severity classification by using recording-based Peripheral Oxygen Saturation Signal. The major feature of the present invention emphasizes on using a recording-based Peripheral Oxygen Saturation Signal (SpO2 signal) as an input, which is different from the deep learning-based prior art of using segment-based signals as an input to a model, and the segment-based signals has only two classification results, i.e. normal or apnea.

Abstract: The present invention provides a method for OSA (Obstructive Sleep Apnea) severity detection using recording-based electrocardiography (ECG) Signal. The major feature of the present invention emphasizes on using a recording-based ECG Signal as an input, which is different from the deep learning-based prior art of using segment-based signals as an input to a model, and the segment-based signals has only two classification results, i.e. normal or apnea. The present invention provides a method for a model to detect and output directly a value of apnea-hypopnea index (AHI) for the OSA Severity.

Abstract: The present invention provides a caption service system for remote speech recognition, which provides caption service for the hearing impaired. This system includes a speaker and a live broadcast equipment at A, a listener-typist and a computer at B, a hearing impaired and a live screen at C, and an automatic speech recognition (ASR) caption server at D. Connect the live broadcast equipment, the computer, the live screen and the ASR caption server with a network. The speaker's audio is sent to the automatic speech recognition (ASR) caption server to be converted into text, which is corrected by the listener-typist, and then the text caption is sent to the live screen of the hearing impaired together with the speaker's video and audio, so that the hearing impaired can see the text caption spoken by the speaker.

Abstract: The present invention provides a speech recognition system for teaching assistance, which provides caption service for the hearing impaired. This system includes a speaker and a automatic speech recognition (ASR) classroom server, a listener-typist and a computer, a hearing impaired and a live screen, all are in the same classroom. Connect the ASR classroom server, the computer and the live screen with a local area network. The speaker's audio is sent to the ASR classroom server by a microphone for being converted into text caption, and then the text caption is sent to the live screen of the hearing impaired together with the speaker's audio so that the hearing impaired can read the text caption spoken by the speaker. The text caption can be corrected by the listener-typist to make it completely correct.

Abstract: The present invention provides a caption service system for remote speech recognition, which provides caption service for the hearing impaired. This system includes a speaker and a live broadcast equipment at A, a listener-typist and a computer at B, a hearing impaired and a live screen at C, and an automatic speech recognition (ASR) caption server at D. Connect the live broadcast equipment, the computer, the live screen and the ASR caption server with a network. The speaker's audio is sent to the automatic speech recognition (ASR) caption server to be converted into text, which is corrected by the listener-typist, and then the text caption is sent to the live screen of the hearing impaired together with the speaker's video and audio, so that the hearing impaired can see the text caption spoken by the speaker.

Abstract: A screw supplying module includes a main body, a feeding body and a holding module. The main body includes a second portion, a third portion and a hollow part. The holding module includes a screw chamber, a plurality of clamping pieces, an elastic piece and a fixing piece. The screw chamber includes an input portion, an outer ring wall and an inner space. The plurality of clamping pieces are inserted into the plurality of holes respectively. Accordingly, the plurality of clamping pieces are partially located in the inner space and are partially exposed from the holes of the screw chamber. The elastic piece is sleeved on the outer ring wall, and the elastic piece abuts against the part of the plurality of clamping pieces exposed from the holes. The fixing piece is sleeved on the outer ring wall and is configured to fix the elastic piece.

Abstract: The present invention provides a method for updating speech recognition system through air. Client ASR servers connect with a central ASR cloud server through Internet. New version of ASR system is stored in the central ASR cloud server for being selected and downloaded by the client ASR servers for using.

Abstract: The present invention provides a system for selecting a special speech recognition model through a general model of an AI speech recognition system for users to select an appropriate model. In addition to the AI speech recognition server of a general model, the present invention additionally prepares speech models in various fields, such as sports event model, financial news model, and game live model. Different users can choose different speech models according to their needs or fields, and they can get better services respectively. If the different users have no special choice, the AI speech recognition server of the general model provides speech recognition services for the different users.

Abstract: The present invention provides a method for generating caption file through URL of an AV platform. By using various websites (such as YouTube, Instagram, Facebook, Twitter) for being inputted with the URL of a desired AV Platform and downloading a required AV file and inputting to an ASR (Automatic Speech Recognition) server according to the present invention. A speech recognition system in the ASR server can abstract an audio file from the AV file for a system operation to get a required caption file. Artificial Neural Networks are used in the present invention.

Abstract: A robot mechanism is provided, including a base, a main body, a motor, a driver, a bottom plate, a flexible heat conductive member, and a controller. The main body is connected to the base and has a housing. The motor and the driver are disposed in the main body, and the driver is electrically connected to the motor. The bottom plate is disposed on the housing and situated between the driver and the housing, and a gap is formed between the bottom plate and the driver. The flexible heat conductive member is disposed between the driver and the housing. The flexible heat conductive member contacts the driver. The controller is detachably disposed in the base.

Abstract: A screw supplying module includes a main body, a feeding body and a holding module. The main body includes a second portion, a third portion and a hollow part. The holding module includes a screw chamber, a plurality of clamping pieces, an elastic piece and a fixing piece. The screw chamber includes an input portion, an outer ring wall and an inner space. The plurality of clamping pieces are inserted into the plurality of holes respectively. Accordingly, the plurality of clamping pieces are partially located in the inner space and are partially exposed from the holes of the screw chamber. The elastic piece is sleeved on the outer ring wall, and the elastic piece abuts against the part of the plurality of clamping pieces exposed from the holes. The fixing piece is sleeved on the outer ring wall and is configured to fix the elastic piece.

Abstract: The present invention provides a method fir emergency treatment by artificial intelligence. An artificial neural network is used as the artificial intelligence. Firstly the artificial neural network is trained to make injury classification, inspection list and medical material scheduling correctly. For a patient entering the hospital, the artificial neural network that has been successfully trained is used to accept a plurality of word vectors and various physiological information of the patient to generate an injury classification. The artificial neural network then determines whether the patient has to perform various inspection items respectively with the highest level of the injury classification. The artificial neural network then determines whether the patient needs the various medical materials with the highest level of the injury classification.

Abstract: An automatic screw tightening module includes a plate assembly, an input module, a screwdriver module, a transmission module, a movable module, an elastic element and a position sensor. The screwdriver module includes a screwdriver and a screwdriver sleeve. The transmission module is connected with an input terminal of the input module and the screwdriver sleeve for allowing the input terminal, the transmission module and the screwdriver sleeve to be rotated synchronously. The movable module is movably disposed on a base plate of the plate assembly. The movable module includes a bearing, and portion of the screwdriver sleeve is accommodated in the bearing, so that the screwdriver module and the movable module are moved relative to the base plate. The elastic element is disposed on the base plate and connected with the movable module. The position sensor is disposed on the base plate for sensing a displacement of the movable module.

Abstract: A robot mechanism is provided, including a base, a main body, a motor, a driver, a bottom plate, a flexible heat conductive member, and a controller. The main body is connected to the base and has a housing. The motor and the driver are disposed in the main body, and the driver is electrically connected to the motor. The bottom plate is disposed on the housing and situated between the driver and the housing, and a gap is formed between the bottom plate and the driver. The flexible heat conductive member is disposed between the driver and the housing. The flexible heat conductive member contacts the driver. The controller is detachably disposed in the base.

Abstract: The present invention provides a method for NAT traversal in VPN so that the VPN can detect the rule of port allocation for NAT outside the VPN to achieve NAT traversal. The communication structure according to the present invention includes a public network, a client network, a destination network, a first NAT, a second NAT. A DNAT-T proxy server is installed between the first NAT and the second NAT and has the function for the VPN to conduct a plurality of (N times) registrations before sending data out to detect the rule for NAT port allocation of the DNAT-T proxy server, and then inform the next NAT port allocation to the other side of the VPN so as to achieve NAT traversal for the data packets in VPN.

Abstract: The present invention provides a method for NAT traversal in VPN so that the VPN can detect the rule of port allocation for NAT outside the VPN to achieve NAT traversal. The communication structure according to the present invention includes a public network, a client network, a destination network, a first NAT, a second NAT. A DNAT-T proxy server is installed between the first NAT and the second NAT and has the function for the VPN to conduct a plurality of (N times) registrations before sending data out to detect the rule for NAT port allocation of the DNAT-T proxy server, and then inform the next NAT port allocation to the other side of the VPN so as to achieve NAT traversal for the data packets in VPN.

Abstract: The present invention provides a treatment method for doorbell communication in order to save the data bandwidth of video streaming communication. The structure of doorbell communication according to the present invention comprises a doorbell, a camera and a microphone, which are connected with a smart phone and a server of a home by a network. When a user pushes the doorbell, the camera captures the face image of the user, and then sends to the server for face recognition. If the face image of the user matches with one of the face templates stored in the server, then the personal data of the related face template is shown on the screen of the smart phone for conducting audio streaming communication without video streaming communication.

Abstract: The present invention provides a method for householder of mansion to manage entrance by smart phone, so that the householder can manage entrance at outside. When a visitor pushes a doorbell of a mansion door machine, the householder at outside uses a smart phone to conduct audio and video communication with the visitor. If the householder agrees to let the visitor go in, the householder instructs the smart phone to generate a password for the visitor to memorize and let the visitor enter the mansion. The password is also sent to a home door machine of the householder. When the visitor an ives the home door machine of the householder, input the password. If the password inputted by the visitor is the same as the password generated by the smart phone, then the home door will open to let the visitor go in.

Abstract: The present invention provides a method for householder of mansion to manage entrance by smart phone, so that the householder can manage entrance at outside. When a visitor pushes a doorbell of a mansion door machine, the householder at outside uses a smart phone to conduct audio and video communication with the visitor. If the householder agrees to let the visitor go in, the householder instructs the smart phone to generate a password for the visitor to memorize and let the visitor enter the mansion. The password is also sent to a home door machine of the householder. When the visitor an ives the home door machine of the householder, input the password. If the password inputted by the visitor is the same as the password generated by the smart phone, then the home door will open to let the visitor go in.