42 results about "Speaker verification" patented technology

The invention discloses a voiceprint characterization method and system for resisting voice conversion and voice synthesis based on deep learning. Relates to the field of speech recognition and biological feature security. Comprising the steps of 1, acquiring an input voice sample signal in an automatic speaker verification system, and extracting an FBANK feature of the voice sample signal; 2, performing depth feature extraction on the FBANK features by using a convolutional neural network (CNN) to obtain a depth feature vector; 3, processing the depth feature vector by using a recurrent neural network (RNN), and generating a forged identity vector of the voice; and 4, classifying the counterfeit identity vectors by using a linear discriminant analysis (LDA) module, and outputting a judgment result. According to the invention, the accuracy of counterfeit detection can be effectively improved in both clean and noise environments.

Traditional speaker verification systems are vulnerable to voice spoofing attacks, such as voice-replay attack, voice-cloning attack, and cloned-replay attack. To overcome these vulnerabilities, a secure automatic speaker verification system based on a novel sign modified acoustic local ternary pattern (sm-ALTP) features and asymmetric bagging-based classifier-ensemble with enhanced attack vector is presented. The proposed audio representation approach clusters the high and low frequency components in audio frames by normally distributing them against a convex function. Afterwards, the neighborhood statistics are applied to capture the user specific vocal tract information.

The application discloses a speaker diary generation method based on audio-visual fusion clustering, and aims to solve the problem of "who speaks at what time" in a multi-speaker scene. The method is realized through the following steps: first, an overlapping-aware speech segmentation model is used to segment the audio segment, solving the problem of overlapping speech; second, an advanced speaker verification model is used to extract the speaker voiceprint features of each audio segment and a speaker score matrix generated through face tracking and speaker detection; then, through an audio-video joint clustering method, the number of clusters is optimized according to the audio features and visual information, and K-means clustering is used to complete speaker clustering; the experimental results show that the system adopting the method achieves the lowest diary error rate (DER) on the Ego4D validation set.

Provided is an ensemble learning (EL)-based speaker verification method. The method includes: data acquisition and preprocessing; selecting and training a group of basic models, and optimizing model parameters; performing similarity scoring on an acquired pair of speaker feature embedding via the group of basic models; constructing a detection cost function (DCF); generating a weight, and performing weighted fusion on scoring results of the group of basic models based on the weight, to obtain a final ensemble model for speaker verification; based on a near-speaking or far-speaking test scenario of a voice sample, the scenario is distinguished and input into the ensemble model, to obtain a final similarity score after weighted fusion; and determining, based on a threshold, whether there is a same speaker, where if the similarity score is greater than the threshold, it is determined that there is a same speaker.

This invention provides an adaptive margin optimization method, system, and electronic device based on speaker verification. The method includes: inputting speech training data including various speech durations into a speaker verification model; determining the loss function of the speaker verification model; adaptively optimizing the margin parameters of the loss function based on the speech durations in the speech training data and a preset target margin for each speech duration; and training the speaker verification model using the margin parameters of the adaptively optimized loss function to determine the acceptable training difficulty of the speaker verification model. This invention utilizes training speech of varying lengths to better simulate real-life scenarios. Through adaptive optimization and fine-tuning of the margin, adjusting the margin according to the duration and similarity of each speech, this method achieves good speaker verification performance for speech of different durations in real-world scenarios.

The application discloses a speaker verification model quantization method, an electronic device and a storage medium, wherein the speaker verification model quantization method comprises the following steps: acquiring real-value weights of all layers of a speaker verification model; and mapping the real-value weights of all layers to a fixed integer set; or dynamically determining binary weights corresponding to the real-value weights of each layer to better match the real-value weight distribution. The method of the application embodiment proposes two new quantization strategies, namely static quantization and adaptive quantization. Furthermore, for static quantization, the application embodiment proposes a weight regularization technique to maintain maximum information entropy and reduce information loss. Furthermore, the application embodiment also proposes an adaptive quantization scheme, which can dynamically determine the best binary of each layer to achieve better alignment with the real-value weight distribution.

A method includes obtaining audio data and identifying an utterance of a wake word or phrase in the audio data. The method also includes generating an embedding vector based on the utterance from the audio data and accessing a set of previously-generated vectors representing previous utterances of the wake word or phrase. The method further includes performing clustering on the embedding vector and the set of previously-generated vectors to identify a cluster including the embedding vector, where the identified cluster is associated with a speaker. The method also includes updating a speaker vector associated with the speaker based on the embedding vector and determining, using a speaker verification model, a similarity score between the updated speaker vector and the embedding vector. In addition, the method includes determining, based on the similarity score, whether a speaker providing the utterance matches the speaker associated with the identified cluster.

The embodiment of the specification provides a generative adversarial network optimization method and system for short speech speaker verification, wherein the method comprises the following steps: acquiring a plurality of pairs of long and short speech acoustic feature samples; inputting the short speech acoustic feature sample into a generator for splicing to obtain a generated pseudo long speech acoustic feature sample; inputting the pseudo long speech acoustic feature sample and the acquired long speech acoustic feature sample into a speaker verification model respectively, outputting a pseudo identity feature sample and a true identity feature sample through the speaker verification model; inputting the true identity feature sample and the pseudo identity feature sample into a discriminator and a classifier, calculating the loss of the discriminator and the classifier through a loss function, and updating the parameters of the discriminator, the classifier and the generator through back propagation optimization. The problem that the discrimination effect of a speaker verification system becomes poor as the speech duration becomes shorter is solved.

This invention discloses a method, apparatus, medium, and device for training a speaker verification model. The method includes: extracting waveforms from acquired speech audio data to obtain audio waveform data corresponding to the speech audio data; inputting the audio waveform data into a preset classification model to output a predicted label corresponding to the speech audio data; determining the loss value of the preset classification model based on the anti-counterfeiting label and the predicted label corresponding to the speech audio data; and adjusting the parameters of the preset classification model using the loss value to obtain a speaker verification model. This invention adopts a supervised learning approach, using real anti-counterfeiting labels to assist in determining the predicted label of the model, and trains the entire model by minimizing the error between the output and the real label, thereby improving training efficiency and reducing costs.

The application discloses a speaker verification method based on a SASFV aggregation model, and relates to the field of speech recognition.The method extracts a log Mel spectrogram through short-time Fourier transform and Mel filtering, generates frame-level features by using an ERes2Net network, introduces a SASFV aggregation model to generate fixed-length speaker-level features in combination with a Fisher Vector variable, a self-attention mechanism and a statistical method, and finally determines the identity of a speaker by using a cosine distance.The application solves the problem that the prior art cannot effectively represent and aggregate features in a short speech task, and significantly improves the accuracy, robustness and performance of a speaker recognition system.

The invention relates to the technical field of voice processing, in particular to a speaker anonymization-oriented stylegan2-based F0 cloning method, which mainly realizes source speaker anonymization protection and comprises the following steps of: preprocessing an original fundamental frequency sequence, and converting a random noise vector into a representation vector of a style space; and a StyleGAN2 architecture is introduced to carry out layer-by-layer feature processing and fusion. A probabilistic fusion strategy is adopted, content features and constant features are added, then layer-by-layer stylization processing is carried out, feature modulation is carried out on each layer by using a style vector, the time sequence resolution is gradually recovered, and the anonymization target of the identity features of the speaker is effectively changed. And finally, combining the generative fundamental frequency (F0), the text information and the anonymized identity features to synthesize anonymized voice so as to hide the identity information of the speaker. And the effectiveness of the speaker anonymization-oriented StyleGAN2-based F0 cloning method is verified through an automatic speaker verification model and an automatic speech recognition system.

A method and an apparatus are disclosed for a speaker verification. The apparatus may comprise at least one processor configured to execute instructions to perform generating, based on an utterance of a speaker to a deep learning-based speaker verification model, a speaker embedding with a preset dimension, verifying, based on the speaker embedding, the speaker, and detecting, based on the verified speaker, an identity of a user associated with the apparatus.

In addition to speaker verification, an LLM continuous conversation check employs a spoken speed check using a phonemes-based spoken speed calculation, and acoustic energy check, and a signal-to-noise estimation to determine whether first and second audio inputs include utterances forming a continuous conversation by the user with the LLM. Results from the various components of the continuous conversation check are fused based on automatically assigned weights in making the determination. Continuous conversation detection for LLMs is therefore more robust, particularly for a very short second utterance. Optionally a distance to microphone check or a semantic consistency check may also be employed.

The application discloses a training method of a multi-modal speaker authentication model and an electronic device, relates to the field of speaker authentication, and comprises the following steps: obtaining an audio-video sequence sample corresponding to a given speaker; performing lip feature coding on each audio frame in the audio-video sequence sample in sequence to determine a corresponding audio sample lip latent vector sequence; performing feature coding on a speaker mouth region in each video frame in the audio-video sequence sample in sequence to determine a corresponding video sample lip latent vector sequence; calculating an audio-video sequence similarity corresponding to the audio sample lip latent vector sequence and the video sample lip latent vector sequence; and training a speaker authentication model according to the audio-video sequence similarity and an audio-video speaker alignment label. Thus, the similarity of the speaker audio and the lip shape graph of the person to be detected is compared by matching the speaker audio and the lip shape graph into the same dimension vector through the latent vector, so that the accuracy of the authentication result can be effectively improved.

The invention provides a speaker verification method and system based on transfer learning, and the method comprises the following steps: constructing a voiceprint feature extraction model, and carrying out the pre-training of the voiceprint feature extraction model through employing out-of-domain annotation data; the voiceprint feature extraction model subjected to pre-training is optimized; the optimized and pre-trained voiceprint feature extraction model is utilized to extract cross-channel voiceprint features of a speaker to be recognized, and cross-channel speaker verification is achieved; according to the method, the problem of feature distribution offset in a cross-channel scene can be effectively solved, and the accuracy and robustness of voiceprint recognition are improved especially in a scene with large-scale out-of-domain annotation data and a small amount of in-domain annotation data given.

The invention provides a speaker feature vector extraction method, a speaker feature vector verification method, a speaker feature vector extraction device, a speaker feature vector verification device and a product, and the speaker feature vector extraction method comprises the steps: processing an obtained input voice signal to obtain a basic acoustic feature; processing the basic acoustic features through at least three speaker attention weight generation modules in sequence; the speaker attention weight module comprises a compressed excitation residual block and a speaker conditional attention module, and the output end of the compressed excitation residual block is connected with the input end of the speaker conditional attention module; the multi-level feature aggregation module outputs a spliced statistical feature vector according to the output features of each speaker conditional attention module; and the full connection layer outputs a speaker feature vector according to the spliced statistical feature vector. The speaker condition attention mechanism is introduced to improve the accuracy of the feature vector of the speaker, and the problem that the verification is not accurate enough in a complex scene such as relatively high speaker similarity is solved when the method is used for speaker verification.

The application provides a kind of robust intelligent synthetic speech speaker confirmation model training method, system, storage medium and electronic equipment, it is related to speech processing technical field.The robust intelligent synthetic speech speaker confirmation model provided by the present application is used to improve the speaker confirmation accuracy of intelligent synthetic speech under strong background noise condition, comprising speech enhancement network, feature extraction network and feature enhancement module.In the training phase of the model, the training set of noisy intelligent synthetic speech data set is transported into speech enhancement network and feature extraction network after STFT feature preprocessing and Fbank feature preprocessing, and the two networks are connected through feature enhancement module for joint training to obtain speaker embedding features with noise-robustness.In the test phase, the model is tested based on the test set of noisy intelligent synthetic speech data set;The optimal model selection is carried out in the mode of repeating the foregoing training and testing alternately until the training iteration number reaches the set maximum value.

The present application relates to the technical field of electronic digital data processing, in particular to a speaker recognition method based on speaker voice micro-motion, comprising: after the voice stream is preprocessed, Fbank features are extracted and sent into a teacher network and a student network respectively to obtain respective corresponding feature embeddings; the feature embeddings obtained by the teacher network and the student network are sent into a loss function and back propagation is performed; the student network is normally iterated, and the teacher network is iterated through an EMA sliding average method; the voiceprint feature information obtained through an ECAPA-TDNN voiceprint model is aggregated and classified with the speaker voice micro-motion information obtained by training the accent data, and speaker recognition is performed; the present application improves the generalization performance of the model by using data enhancement and other methods, avoids fitting in the channel features, does not require manual labeling, and helps the speaker verification model to achieve the ability to identify speakers in a larger population by introducing new features.

A computer-implemented server-performed method for personalising a trained speaker verification machine learning model for voice authentication of specific users obtains a “positive” audio sample of t

The invention discloses a zero sample ASMR generation method and system based on a large language model, and aims to solve the problems that in the prior art, personalized ASMR voice cannot be generated in a zero sample mode, and a high-quality ASMR special data set is lacked. The method comprises the following steps that speaker prompt and task signals of texts to be synthesized and normal or ASMR style voice are obtained; retrieving a matching task cue from the virtual pool of speakers based on the speaker verification system; pre-training the large language model to generate a voice token sequence containing a target style; the stream matching acoustic decoder fuses the voice token sequence and speaker acoustic information to generate a target Mel spectrum; the vocoder synthesizes a target audio; the system comprises an input module, a virtual speaker pool module, a large language model style coding module, a stream matching decoding module and a vocoder module, and further comprises a data set module for storing a DeepASMR-DB data set covering 9 types of themes, Chinese and English and over 670-hour voices. According to the method, zero-sample ASMR generation is realized, the tone and real breath sound of a speaker are reserved, and the audio fidelity is high.

A method for obtaining enrolment data for a speaker verification mode, comprises capturing audio samples of utterances spoken by a speaker, generating an embedding vector for some samples, identifying

Implementations relate to automatic generation of speaker features for each of one or more particular text-dependent speaker verifications (TD-SVs) for a user. Implementations can generate speaker features for a particular TD-SV using instances of audio data that each capture a corresponding spoken utterance of the user during normal non-enrollment interactions with an automated assistant via one or more respective assistant devices. For example, a portion of an instance of audio data can be used in response to: (a) determining that recognized term(s) for the spoken utterance captured by that the portion correspond to the particular TD-SV; and (b) determining that an authentication measure, for the user and for the spoken utterance, satisfies a threshold. Implementations additionally or alternatively relate to utilization of speaker features, for each of one or more particular TD-SVs for a user, in determining whether to authenticate a spoken utterance for the user.

The invention relates to the technical field of speech recognition, in particular to a speaker verification model adversarial training method and system based on general adversarial disturbance, and the method comprises the steps: obtaining an indication label and random Gaussian noise with a fixed length; according to the indication label, converting the random Gaussian noise into general adversarial disturbance by using an adversarial disturbance generator, and adding the general adversarial disturbance to the clean audio sample in different modes to obtain an audio adversarial sample; using a discriminator to obtain a probability value of whether the audio confrontation sample is a target audio; and inputting the audio confrontation sample and the target audio into the speaker verification model to obtain an audio embedding vector, and obtaining similarity loss of the speaker verification model so as to perform confrontation training on the speaker verification model. Therefore, the problem that in confrontation training, confrontation disturbance is high in dependency on samples, confrontation disturbance numerical values are single, and the effect in a physical domain is poor, so that the robustness of a speaker verification model is poor is solved.

The application discloses a speaker verification method and system based on double-flow low-rank adaptive and anti-decoupling, comprising: based on a pre-trained speech network, a double-flow low-rank adaptive anti-decoupling network is constructed, and the original weight parameters of the pre-trained speech network are frozen; based on a language feature extraction branch and a speaker feature extraction branch, original speech data is subjected to feature extraction respectively to obtain language features and speaker features; the language features are input into the shared discriminator to perform language classification prediction to complete language boundary anchoring; after gradient inversion processing of the speaker features, the speaker features are input into the shared discriminator which has completed language boundary anchoring to perform anti-decoupling; identity recognition is performed based on the speaker features subjected to anti-decoupling constraint; corresponding training losses are calculated respectively, and iterative updating is performed based on the training losses. The application can improve the acceptance rate of cross-language speech of the same person and the rejection rate of the same language speech of different persons, and is suitable for high-precision speaker verification in a multi-language environment.

This invention discloses a time-varying speaker template update method based on risk gating, belonging to the field of speaker verification and speech signal processing technology. The method first constructs an initial speaker template and anchor template from registered speech, then verifies the test speech. When the update conditions are met, an attractive update is performed first, then candidate negative samples are obtained. A risk gating quantity is constructed through uncertainty-driven or deterministic-driven gating branches to determine the repulsion strength. Projective repulsion update is then performed in the negative direction, followed by step size pruning and anchor point retraction. This method can reduce template drift caused by erroneous updates and improve the stability and discriminative power of time-varying speaker verification.

The invention discloses a multi-channel voice processing method and system, and belongs to the field of artificial intelligence and voice signal processing. Comprising the following steps: acquiring a multi-channel audio signal and constructing a single-channel pre-training model of SSL; based on the multi-channel audio signal, performing structure optimization on the single-channel pre-training model of the SSL to obtain a multi-channel voiceprint feature extraction pre-training model; performing multi-stage joint optimization training, performing fine tuning on the multi-channel voiceprint feature extraction pre-training model in combination with AAM loss, and generating a multi-channel voice processing model; and when a to-be-processed multi-channel audio signal is received, processing the to-be-processed multi-channel audio signal through the multi-channel voice processing model, and outputting a high-discrimination multi-channel voiceprint feature. The invention aims to improve the accuracy and robustness of speaker verification in a multi-channel scene, remarkably reduce the error rate and improve the adaptability of a system to a complex acoustic environment.

The voice synthesis method, voice synthesis device, electronic equipment and storage medium provided by the present application relate to the fields of artificial intelligence and financial technology. The method comprises the following steps: extracting speaker features from sample voice data sets through a feature embedder in an initial speaker verification model to obtain sample set speaker features; classifying the sample set speaker features through a feature classifier in the initial speaker verification model to obtain a sample verification speaker category, then adjusting parameters of the initial speaker verification model to obtain a target speaker verification model; extracting speaker features from target voice data through a feature embedder in the target speaker verification model to obtain target speaker features; and generating voice according to the target speaker features and target text features to obtain target synthesized voice data. The present application can alleviate the adverse effects of background noise in voice data and improve the accuracy of voice synthesis.

The present application relates to the technical field of voiceprint recognition, and particularly relates to a text-independent speaker verification method based on identity information and semantic information disentanglement. The technical scheme of the present application is completed by building a neural network, training a model, and testing a result. sch The semantic content is represented by a fixed-length feature f spk The correlation between the voiceprint feature f sch and the semantic feature f sch is calculated using a disentanglement method such as mutual information to achieve disentanglement between the voiceprint feature and the semantic feature, obtain more accurate and robust speaker voiceprint features, and solve the problem of poor system generalization performance caused by interference of semantic content factors in the data set extracted by the speaker recognition system.

The application discloses a far-field speaker verification method based on self-distillation pre-training and meta-learning fine-tuning, and the process is as follows: log mel spectrum features are extracted from near-field training speech as the input of the network, and a self-distillation learning method is used to pre-train a deep neural network; then, log mel spectrum features of far-field training speech are used as the input of the network, and a meta-learning method is used to fine-tune the pre-trained network; then, log mel spectrum features of the registered speech are input into the pre-trained and fine-tuned network to obtain a transformed feature vector, and log mel spectrum features of the test speech are input into the fine-tuned and pre-trained network to obtain a transformed feature vector; finally, the distance between the transformed feature vector of the test speech and the transformed feature vector of the registered speech is calculated, and it is judged whether the two are from the same speaker. The method improves the network training efficiency and improves the speaker verification accuracy when the recording environment of the registered speech and the test speech is inconsistent.

A method and an apparatus are disclosed for a speaker verification. The apparatus may comprise at least one processor configured to execute instructions to perform generating, based on an utterance of a speaker to a deep learning-based speaker verification model, a speaker embedding with a preset dimension, verifying, based on the speaker embedding, the speaker, and detecting, based on the verified speaker, an identity of a user associated with the apparatus.