35 results about "Normal speech" patented technology

Provided is an information processing device that performs processing related to voice input.The information processing device includes: a classification unit that classifies an uttered voice into a normal voice and a whisper on the basis of a voice feature amount; a recognition unit that recognizes a whisper classified by the classification unit; and a control unit that controls processing based on a recognition result of the recognition unit. The information processing device further includes a normal voice recognition unit that recognizes a normal voice classified by the classification unit, in which the control unit performs processing corresponding to a recognition result of a whisper by the recognition unit on a recognition result of the normal voice recognition unit.

The invention provides a multi-mode dysarthria speech reconstruction system and a multi-mode dysarthria speech reconstruction method. The system comprises an encoder and a decoder. The encoder is configured to: extract a multi-modal feature from a multi-modal input of dysarthria speech and generate phoneme embedding based on the multi-modal feature; predicting a sequence of phoneme durations based on the phoneme embedding and generating an extended phoneme embedding and a predicted tone sequence; and extracting a speaker identity characterization of the dysarthria speech speaker. The decoder is configured to reconstruct a normal speech waveform corresponding to dysarthria speech with the extended phoneme embedding, the predicted tone sequence, and the speaker identity representation as inputs.

The invention discloses a voice interaction application testing method and device and related equipment. The method comprises the steps of obtaining a target file stream from a preset file, wherein the preset file comprises a file stream generated after a normal voice interaction event stream of a voice interaction application is serialized; deserializing the target file stream to generate various voice type events; distributing the various voice type events to a voice interaction application; and driving the voice interaction application to execute the various voice type events to obtain a test result of the voice interaction application. According to the embodiment of the invention, the method does not involve the contact of an operation interface, is suitable for the test of a non-touchable voice interaction application, and guarantees the normal operation of the voice interaction application in the vehicle-mounted terminal.

The present application belongs to the technical field of intelligent teaching, and particularly relates to a teaching quality evaluation method and system based on deep learning. The method comprises extracting feature data to be evaluated from normal speech, evaluating the feature data to be evaluated by using a speech evaluation model to generate an evaluation result, and determining the quality grade of the normal speech according to the evaluation result, wherein the feature extraction from noise speech and noise speech comprises extracting amplitude information and frequency information from a sound production section. The present application performs screening on teaching speech to identify abnormal sound sections; through voiceprint comparison, the teaching speech in which the abnormal sound sections that can match the pre-stored voiceprint are classified as noise speech, and the teaching speech that cannot be matched is classified as noise speech, so as to distinguish the noise speech originating from the background environment from the noise speech originating from the teaching subject, overcome the evaluation error problem caused by regarding the two as noise without distinction, and lay a data foundation for subsequent evaluation.

The invention belongs to the field of natural language processing, and relates to a hatred recognition method combining multi-layer hatred generation and progressive comparative learning, which comprises the following steps: acquiring a private hatred recognition model trained by speech input to obtain speech characteristics, and obtaining a hatred recognition result according to the speech characteristics; the training process of the hatred recognition model comprises the following steps: acquiring a speech set, and inputting the speech set into a feature coding module to obtain a speech coding feature fi; the speech coding characteristics comprise hatred speech coding characteristics fih and normal speech coding characteristics fin; inputting the characteristic fih into a multi-layer hatred generation module to obtain a multi-layer hatred speech Xih; inputting the speech Xih into a feature extraction module to obtain a multi-layer hatred speech feature Fih; calculating a total loss function value according to the features Fih and fi, and updating model parameters according to the total loss function value until a trained model is obtained; according to the method, different layers of hatred expression modes are fully covered through the multi-layer hatred generation module, so that various expressions can be accurately recognized, and the recognition robustness is improved.

The invention provides an interphone noise suppression method and a noise suppression device, which are used for judging whether a sound signal is noise or normal voice by utilizing a decibel value of an actually acquired sound signal and / or a ratio of a voice signal to a total sound signal, and switching on or switching off an information channel between an interphone and an earphone according to a judgment result. Moreover, the decibel threshold value of the normal voice signal is dynamically updated according to the environmental data, so that the accuracy of noise discrimination and the noise suppression effect are improved. The noise suppression device designed by the invention is externally hung between the interphone and the earphone, when the noise suppression device is used for suppressing noise, the interior of the original interphone does not need to be disassembled or maintained, and the noise suppression device is particularly suitable for maintenance of old equipment interphones.

This invention relates to the field of speech recognition technology, specifically to a method for suppressing false triggering in a speech recognition system. This method acquires and segments candidate wake-up speech signals, performs audible acoustic analysis and high-frequency carrier anomaly analysis on each speech segment to be verified, and calculates a command validity score by combining cross-segment consistency, audible injection risk, and target secondary discrimination threshold. Based on this, false triggering speech signals are suppressed. This invention combines candidate recognition, segment verification, audible acoustic analysis, and high-frequency carrier anomaly analysis, and comprehensively considers cross-segment physical consistency, audible injection risk, and target secondary discrimination threshold to generate a command validity score, thereby improving the accuracy, security, and reliability of false triggering recognition and normal speech command response.

The invention relates to the technical field of speech recognition, and discloses a phoneme recognition model training method, a phoneme recognition method and a phoneme recognition device.A first sample phoneme sequence corresponding to normal speech data is recognized through a first phoneme recognition model, and a second sample phoneme sequence corresponding to dysarthria speech data is recognized through a second phoneme recognition model; and taking the second sample phoneme sequence as a reference phoneme sequence in a scene without a fixed text. According to the method, fine tuning training is performed on the basis of the first phoneme recognition model according to the individual dysarthria data of the dysarthria patient to obtain the individualized second phoneme recognition model, and the situation of free dialogue or poor text compliance can be covered even in a scene without a fixed text, so that the accuracy of recognition of the dysarthria is improved. The actual pronunciation ability of the patient can be reflected in a natural communication context, dependence on fixed texts and artificial phoneme level alignment labeling is avoided, higher flexibility and adaptability are achieved, and a reliable basis can be provided for rehabilitation training and curative effect tracking.

The present invention discloses a speech recognition method for an air traffic control speech recognition model based on speech rate perception, comprising: step 1: counting the normal speech rate range of air traffic control speech; step 2: constructing an air traffic control speech recognition model including the speech rate estimation layer; step 3: defining a joint loss function, training the air traffic control speech recognition model, and dynamically adjusting the CTC loss weight coefficient and the speech rate perception loss weight coefficient in the joint loss function during the training process; step 4: using the trained speech recognition model to recognize air traffic control speech data. The present invention introduces speech rate perception loss constraints to form a multi-task learning mechanism that combines the main task of speech recognition with the auxiliary task of speech rate regularity learning, which significantly improves the robustness and recognition accuracy of the air traffic control land-to-air call speech recognition model in complex environments, and provides more reliable technical support for the safe and efficient operation of air traffic control.

The invention relates to the technical field of audio processing, and discloses a spoken language evaluation method, the method is applied to spoken language evaluation equipment integrated with a speech speed prediction model, and the method comprises the following steps: obtaining a frame level acoustic feature of a to-be-detected audio, and calculating a speech speed value of the to-be-detected audio according to the frame level acoustic feature; inputting the frame-level acoustic features of the to-be-tested audio and the speech speed value of the to-be-tested audio into a speech speed prediction model to obtain a speech speed adjustment value of the to-be-tested audio; adjusting the speech speed of the to-be-tested audio according to the speech speed adjustment value of the to-be-tested audio; and scoring based on the to-be-tested audio after the speech speed is adjusted. After the speech speed of the audio to be detected is adjusted, the speech speed is more consistent with the normal speech speed, so that the start frame and the end frame of each audio clip corresponding to each phoneme of the reference text in the audio to be detected can be more accurately detected in the alignment process. Therefore, the alignment effect of the to-be-tested audio and the reference text is improved, and the scoring accuracy is improved. The invention further discloses a spoken language evaluation device and equipment and a storage medium.

To provide an abnormal state detection system, an abnormal state detection method, and an abnormal state detection program capable of detecting the abnormal state of a vehicle without preparing scream data in advance.SOLUTION: An abnormal state detection system comprises: a voice abnormality detection unit that includes a model generation unit for generating a model of voice data of a person in a vehicle under normal circumstances as a normal-state model from voice data including the voice of the person, and detects abnormalities in the voice based on the current voice data and the normal-state model of the person; a vehicle body abnormality detection unit that detects abnormalities in the body of the vehicle based on acceleration data of the vehicle; and an abnormal state determination unit that determines whether or not the vehicle is in an abnormal state based on the detection result of the voice abnormality detection unit and the detection result of the vehicle body abnormality detection unit.SELECTED DRAWING: Figure 2

Embodiments of the present application disclose a speech processing method and an electronic device, and relate to the technical field of information processing. The method comprises: obtaining original speech; preprocessing the original speech to obtain a plurality of original speech features; in the case where the plurality of original speech features comprise a first speech feature, extracting a voiceprint feature from the first speech feature to obtain a first voiceprint feature, the speech type corresponding to the first speech feature being a whisper speech type; determining a target voiceprint feature corresponding to the first voiceprint feature from a target voiceprint feature library; and performing whisper speech conversion on the first speech feature based on the target voiceprint feature to obtain converted normal speech. According to the present application, the same tone as the real tone of the user or the tone specified by the user can be restored in the whisper speech conversion process, thereby improving the user experience.

The invention relates to the technical field of voice information security, and discloses a recording interference prevention method and device, and the method comprises an environment perception step, an equipment recognition and positioning step, an interference strategy decision-making step, a dynamic sequence generation step, a signal fusion construction step, a parametric array modulation step, a directional emission step, and an evaluation adjustment step. The device corresponds to the method. According to the application, different recording devices are accurately distinguished through an acoustic fingerprint identification technology, and a targeted interference strategy can be generated according to specific acoustic features of the devices; by combining an acoustic parametric array and a beam forming technology, accurate directional transmission of interference sound waves is realized, interference energy is enabled to act on target equipment in a concentrated manner, and the interference efficiency is improved; through fusion of multiple dynamic sequences and construction of composite baseband signals, the unpredictability and anti-filtering capability of interference signals are enhanced. Therefore, the high-efficiency interference performance is maintained, the influence on normal voice communication is reduced, and accurate and self-adaptive anti-recording protection is realized.

A computer may train a single-class machine learning using normal speech recordings. The machine learning model or any other model may estimate the normal range of parameters of a physical speech production model based on the normal speech recordings. For example, the computer may use a source-filter model of speech production, where voiced speech is represented by a pulse train and unvoiced speech by a random noise and a combination of the pulse train and the random noise is passed through an auto-regressive filter that emulates the human vocal tract. The computer leverages the fact that intentional modification of human voice introduces errors to source-filter model or any other physical model of speech production. The computer may identify anomalies in the physical model to generate a voice modification score for an audio signal. The voice modification score may indicate a degree of abnormality of human voice in the audio signal.

The utility model discloses an elevator control device which comprises an elevator control panel body and a plurality of floor keys arranged on the front surface of the elevator control panel body, and further comprises a sound hole square plate with a plurality of through holes formed in the surface, and a rectangular plate groove corresponding to the sound hole square plate is formed in the front surface of the elevator control panel body. A dismounting mechanism is arranged between the sound hole square plate and the elevator control panel main body; according to the elevator control panel, the elevator control panel in the prior art is improved and optimized, and the sound hole square plate convenient to disassemble and assemble is designed on the basis of the original structure, so that the sound hole square plate can be quickly disassembled for maintenance and cleaning through the disassembling and assembling mechanism when normal voice broadcast is influenced by dust accumulated in the through hole in the sound hole square plate subsequently, and the working efficiency is improved. And the whole elevator control panel main body does not need to be detached from the inner wall of the elevator car, so that the maintenance convenience of the whole device is greatly improved.

The application discloses an anti-adversarial audio defense method and system based on audio conversion and transcription difference detection, and belongs to the technical field of artificial intelligence security. The method comprises the following steps: step 1, standardizing and multi-dimensionally converting an input original audio; step 2, calculating a character error rate between the original audio and an audio sample transcription result, and comparing the character error rate with a pre-set detection threshold; and step 3, triggering a security response mechanism when the audio contains potential adversarial interference, and submitting the audio to an automatic speech recognition model for transcription processing when the audio is normal speech. The application can realize rapid detection and interception of adversarial audio samples without modifying an existing automatic speech recognition model, significantly improves the stability and robustness of detection, enables the system to realize real-time detection with low delay, balances high accuracy and low false alarm rate, and meets the actual needs of a voice interaction scene.

The invention provides a multi-tone-area voice acquisition method and device, electronic equipment, a chip and a medium, and relates to the technical field of voice interaction. The method comprises the steps that in response to a first state change signal of a first seat where a target object is located, a target sound area corresponding to the first seat is updated, and the first state change signal comprises at least one of a position change signal and a posture change signal and a pressure change signal; and collecting a voice signal of the target object in the target voice area. Through the technical scheme provided by the invention, the problem that normal voice interaction is influenced when a driver adjusts the seat in the cabin is solved, and the user experience of voice interaction in the cabin is improved.

The embodiment of the invention provides a vehicle-mounted voice protection method, device and equipment and a computer readable storage medium. The method comprises the following steps: acquiring an in-vehicle voice signal in real time; analyzing the voice signal to obtain power spectrum distribution characteristics; according to the power spectrum distribution characteristics, multiple paths of periodic short pulse sequences are constructed; generating a multi-channel ultrasonic masking signal according to the multi-path periodic short pulse sequence; and transmitting the multi-channel ultrasonic masking signal by using an ultrasonic phased array, and forming an ultrasonic sound field in a target area in the vehicle. In this way, the directional interference characteristic of the ultrasonic phased array can be utilized, accurate noise injection of illegal recording equipment is achieved, and therefore vehicle-mounted voice privacy is effectively protected under the condition that normal voice communication is not affected.

Provided is an information processing device for performing a process pertaining to voice conversion in which pitch information such as low-sound voice and hoarseness voice is not normally generated. The information processing apparatus includes: a voice unit conversion unit for generating a voice unit from a voice waveform; and a unit voice conversion unit for reconstructing a voice waveform from the voice unit. A unit speech conversion unit is trained in advance using self-supervised learning of a masked language model type without a text marker accompanied by a specific speaker using normal speech and low-tone speech so as to generate a speech unit that is common to the normal speech and the low-tone speech. And the unit of speech is used as a potential representation of the absorption of the difference between normal speech and low speech.

The invention relates to the technical field of voice verification, in particular to a voice detection method based on Fourier transform feature extraction of a logarithmic graph, which comprises the following steps of: constructing a translation operator of a voice graph, describing attenuation of a dependency relationship between voice samples by using an exponential function, generating a Laplacian matrix of the graph, and extracting the Fourier transform feature of the logarithmic graph. And representing the voice signal as an undirected graph to capture an intra-frame and inter-frame structural relationship; the method comprises the following steps: mapping a sample value of a voice signal into a graph node signal, converting the voice signal from a time domain to a graph frequency domain, extracting frequency domain features, and forming enhanced feature representation by synchronously combining intra-frame and inter-frame oscillation analysis and combining time domain features; and generating a detection score to judge whether the voice signal has a playback attack or belongs to normal voice. According to the invention, by introducing logarithmic graph Fourier transform and graph signal processing methods, the limitation of the prior art in playback voice detection is effectively solved, and the comprehensiveness and discrimination capability of feature extraction and the performance of a detection system are remarkably improved.

The invention provides audio processing for assisting communication between a plurality of users of an online video game. The system is designed for when a first user is unable to use normal speech vi

This invention provides a noise reduction and sound pickup device and method. The noise reduction and sound pickup device includes at least one noise pickup, a target sound pickup device, and a sound signal processing unit. The at least one noise pickup is positioned close to a corresponding noise source, suitable for acquiring ambient noise emitted by the noise source and generating an ambient noise frequency signal. The target sound pickup device is positioned close to a target sound source, suitable for acquiring the target sound from the target sound source and generating a target sound frequency signal. The sound signal processing unit inverts and amplifies at least a portion of the frequency signal in the ambient noise frequency signal, superimposes the inverted and amplified ambient noise signal with the target sound signal to obtain a useful sound, and outputs the useful sound. The noise reduction and sound pickup device and method can acquire and eliminate various complex noises emitted by large equipment, while simultaneously preserving and amplifying the required sound signal to ensure normal voice communication and improve communication quality.

This application relates to a method for automatically switching between mesh calls and 5G data network calls. Specifically a Bluetooth device constructs a mesh network to form a mesh group One of the nodes is used as the master node, and an online loop is created on the AP P. All Bluetooth devices communicate with the application in a two-way dynamic heartbeat data mode, and the cloud synchronizes information in the virtual network formed by mapping the nodes; The voice data between online nodes is communicated in mesh mode; when the nodes of Bluetooth devices are offline or reconnected, the node device automatically switches to the data network mode through the application and performs network communication using the data network mode via the cloud. This automatic switching method between mesh calls and 5G data network calls monitors the online status of the device in real time and automatically switches the voice data transmission to the cloud network when the device / node goes offline, thereby ensuring the normal voice data transmission and reception of non - communicating nodes and ensuring the real - time nature, continuity and completeness of voice data transmission. ​

Provided is an information processing device that perform processing related to speech conversion of a speech that is not normally uttered and does not include pitch information such as a whisper or a faint speech. The information processing device includes a speech-to-unit encoder that generates an acoustic unit from a speech waveform, and a unit-to-speech decoder that reconstructs a speech waveform from an acoustic unit. The unit-to-speech decoder is subjected to preliminary learning by self-supervised learning of a Masked Language Model type using a normal speech and a whisper without a text label of a specific speaker to generate an acoustic unit common to the normal speech and the whisper, the acoustic unit being a latent expression in which a difference between the normal speech and the whisper is absorbed.

The present invention provides a method and system for voice noise analysis. The method includes: obtaining voice data to be analyzed; extracting noise audio segments containing only noise from the voice data to be analyzed; determining the noise intensity level corresponding to each noise audio segment based on the noise intensity index of each noise audio segment and a preset noise intensity division level; and determining the noise level evaluation result of the voice data to be analyzed according to the distribution of the noise intensity levels corresponding to each noise audio segment. By calculating the noise intensity index of the noise audio segments containing only noise for separate analysis, and then determining the noise level evaluation result of the entire voice data to be analyzed according to the distribution of the noise intensity levels of all the noise audio segments, the influence of normal speech is avoided, an objective evaluation of the noise level of the voice data to be analyzed is achieved, and there is no need to refer to an audio, so the application range is wider and the noise situation in various scenarios can be accurately reflected.

A voice wake-up interactive response method and system are disclosed, primarily used to detect whether a user has spoken after a voice interaction system has been woken up. The given time window after wake-up is divided into different regions, and different techniques are used to process each region. Specifically, in the false trigger region immediately following the wake-up word detection, the system detects whether the user has spoken and determines whether the spoken words are confused with the ending sound of the wake-up word. In the normal speech detection region following the false trigger region, the system detects whether the user has spoken. In the blind zone detection region following the normal speech detection region, the system detects whether the user has spoken. If no spoken words are detected in the blind zone detection region, the system calculates the fundamental frequency. If the fundamental frequency calculation indicates the presence of spoken words for a certain duration, then spoken words are considered detected. Using the method and system of this invention can significantly reduce the proportion of false detections and false negatives.

The invention discloses a pseudo-ear language generation method, device and equipment. The pseudo-ear language generation method comprises the following steps: acquiring normal voice; extracting multi-scale acoustic features of the normal voice; based on the multi-scale acoustic features, determining sounding effort capable of representing sounding force and clearness; and performing joint modulation of the cross-domain acoustic parameters based on the sounding effort, wherein the joint modulation comprises at least two of a time domain, a frequency domain and an excitation domain. The acoustic parameters of the time domain are adjusted based on the sounding effort, so that dynamic fluctuation caused by semantic key points, emotion emphasis or breathing rhythm in real ear language can be reflected; the acoustic parameters of the frequency domain are adjusted based on the sounding effort, so that the phenomenon that the sounding is clearer by force can be simulated; the acoustic parameters of the excitation domain are adjusted based on the sounding effort, and the effect of enhancing the high-frequency hoarseness sound during forced blowing can be simulated; therefore, the naturalness and interpretability of the pseudo-ear language can be effectively improved.

The invention discloses a speech recognition method of an air traffic control speech recognition model based on speech speed perception. The speech recognition method comprises the following steps: step 1, counting a normal speech speed interval of air traffic control speech; 2, constructing an air traffic control speech recognition model containing the speech speed estimation layer; 3, defining a joint loss function, training an air traffic control speech recognition model, and dynamically adjusting a CTC loss weight coefficient and a speech speed perception loss weight coefficient in the joint loss function in the training process; and step 4, using the trained voice recognition model to recognize the air traffic control voice data. The speech speed perception loss constraint is introduced to form a multi-task learning mechanism combining a speech recognition main task and a speech speed rule learning auxiliary task, the robustness and the recognition accuracy of an air traffic control air-ground communication speech recognition model in a complex environment are remarkably improved, and a more reliable technical support is provided for safe and efficient operation of air traffic control.

This invention relates to the field of speech verification technology, and more particularly to a speech detection method based on logarithmic graph Fourier transform feature extraction, comprising the following steps: constructing a translation operator for the speech graph, using an exponential function to describe the decay of dependencies between speech samples, generating the Laplacian matrix of the graph, and representing the speech signal as an undirected graph to capture intra-frame and inter-frame structural relationships; mapping the sample values ​​of the speech signal to graph node signals, transforming the speech signal from the time domain to the graph frequency domain, extracting frequency domain features, and forming an enhanced feature representation by synchronously merging intra-frame and inter-frame oscillation analysis and combining time domain features; generating a detection score to determine whether the speech signal is a playback attack or belongs to normal speech. This invention, by introducing logarithmic graph Fourier transform and graph signal processing methods, effectively solves the limitations of existing technologies in playback speech detection, significantly improving the comprehensiveness of feature extraction, discriminative ability, and performance of the detection system.

The invention relates to the technical field of intelligent household appliances, and discloses a method for controlling an air conditioner, which comprises the following steps: when a do-not-disturb mode takes effect, determining a current effective sub-mode; when the sub-mode is an insensitive mode, checking the wake-up information through the local and the cloud, and judging whether normal voice interaction is started or not; and when the sub-mode is the no-disturbing mode, starting mute voice interaction. Through the insensitive mode, the air conditioner awakening sensitivity is reduced, air conditioner voice interference is reduced, timely voice interaction can be conducted in response to awakening of the air conditioner by the user, and the user can control the air conditioner conveniently. The mute voice interaction is started in the non-disturbing mode, even if the user wakes up the air conditioner, no sound is made, and the requirement of the user for mute control is met. By setting the insensitive mode and the no-disturbing mode, the increasing control requirement of the user on the no-disturbing mode of the air conditioner is met, and the user experience is improved. The invention further discloses a device for controlling the air conditioner, the air conditioner and a computer readable storage medium.