Voice Print Research Articles

SiO2-capped ZnO quantum dot based highly sensitive optical fiber humidity sensor with potential applications in human breath monitoring and voice print recognition

This article describes the development and characterization of an optical fiber humidity sensor employing intensity modulation via the evanescent wave (EW) absorption technique. For the development of the sensor, SiO2-capped ZnO quantum dot (QDs) thin film is synthesized over the decladded portion of plastic cladding silica (PCS) fiber via the sol-gel method. A thorough experimental investigation was conducted by varying the thickness of the sensing film to optimize the sensor’s response. The sensing probe with optimized film thickness of 891 nm demonstrates a linear response over 30.5%–92.5%RH with an enhanced sensitivity of 46.2 mV/%RH (0.0138RH−1). Very fast response and recovery times of 2 s and 2.5 s are observed during humidification and dehumidification for the optimized sensing probe. The maximum resolution recorded during the short stability test is ±0.12%RH. Additionally, the proposed sensor demonstrates a very high degree of repeatability, reversibility, and stability. The proposed sensor has also been tested for human breath monitoring and voice print recognition. The result shows the sensor is able to detect minute humidity fluctuations in exhaled air during breathing and speaking.

ATC-SD Net: Radiotelephone Communications Speaker Diarization Network

This study addresses the challenges that high-noise environments and complex multi-speaker scenarios present in civil aviation radio communications. A novel radiotelephone communications speaker diffraction network is developed specifically for these circumstances. To improve the precision of the speaker diarization network, three core modules are designed: voice activity detection (VAD), end-to-end speaker separation for air–ground communication (EESS), and probabilistic knowledge-based text clustering (PKTC). First, the VAD module uses attention mechanisms to separate silence from irrelevant noise, resulting in pure dialogue commands. Subsequently, the EESS module distinguishes between controllers and pilots by levying voice print differences, resulting in effective speaker segmentation. Finally, the PKTC module addresses the issue of pilot voice print ambiguity using text clustering, introducing a novel flight prior knowledge-based text-related clustering model. To achieve robust speaker diarization in multi-pilot scenarios, this model uses prior knowledge-based graph construction, radar data-based graph correction, and probabilistic optimization. This study also includes the development of the specialized ATCSPEECH dataset, which demonstrates significant performance improvements over both the AMI and ATCO2 PROJECT datasets.

Optimizing feature fusion for improved zero-shot adaptation in text-to-speech synthesis

In the era of advanced text-to-speech (TTS) systems capable of generating high-fidelity, human-like speech by referring a reference speech, voice cloning (VC), or zero-shot TTS (ZS-TTS), stands out as an important subtask. A primary challenge in VC is maintaining speech quality and speaker similarity with limited reference data for a specific speaker. However, existing VC systems often rely on naive combinations of embedded speaker vectors for speaker control, which compromises the capture of speaking style, voice print, and semantic accuracy. To overcome this, we introduce the Two-branch Speaker Control Module (TSCM), a novel and highly adaptable voice cloning module designed to precisely processing speaker or style control for a target speaker. Our method uses an advanced fusion of local-level features from a Gated Convolutional Network (GCN) and utterance-level features from a gated recurrent unit (GRU) to enhance speaker control. We demonstrate the effectiveness of TSCM by integrating it into advanced TTS systems like FastSpeech 2 and VITS architectures, significantly optimizing their performance. Experimental results show that TSCM enables accurate voice cloning for a target speaker with minimal data through both zero-shot or few-shot fine-tuning of pretrained TTS models. Furthermore, our TSCM-based VITS (TSCM-VITS) showcases superior performance in zero-shot scenarios compared to existing state-of-the-art VC systems, even with basic dataset configurations. Our method’s superiority is validated through comprehensive subjective and objective evaluations. A demonstration of our system is available at https://great-research.github.io/tsct-tts-demo/, providing practical insights into its application and effectiveness.

Technique and result test of transformer voire feature extraction under high noise environment

In the whole power system, the importance of the transformer is self-evident, if its fault is bound to cause adverse effects on the whole power system. However, the previous fault diagnosis method is not only time-consuming and laborious, but also has great danger, once it is not handled properly, it may also pose a threat to the life safety of operators. At this time, the transformer fault diagnosis mode based on voice print follows and becomes a big tool for fault diagnosis. However, the surrounding environment of the transformer is changeable and noisy, which will cause interference to the voice print signal and affect the fault diagnosis result. In this regard, this paper will start from this key point and solve this unfavorable factor by using different denoising technologies and feature extraction technologies. The final empirical results show that the improved wavelet threshold technology has the best denoising ability, and the feature parameters obtained under the transformer sound frequency cepstrum coefficient (TFCC) feature extraction technology can better improve the fault diagnosis accuracy. It can solve the problem of low fault diagnosis rate of transformer in different environments, and can be used in practical engineering.

Deepfake Speech Recognition and Detection

Deepfake technology, especially deep voice, which has been derived from artificial intelligence in recent years, is potentially harmful, and the public is not yet wary. However, many speech synthesis models measure the degree of true restitution by Mean Opinion Rating (MOS), a subjective assessment of naturalness and quality of speech by human subjects, but in future it will be difficult to distinguish the interlocutor’s identity through the screen. For this reason, this study addresses the threat posed by this new technology by combining representational learning and 0transfer learning in two sub-systems: a recognition system and a voice print system. The recognition system is responsible for the detection of which voice is a fake voice generated by speech conversion or speech synthesis techniques, while the acoustic system is responsible for the verification of the speaker’s identity through acoustic features. In the speech recognition system, we use the representation learning method and the transfer classification method. We use X-vector data for training, and then fine-tune the model using four types of marker data to learn the representation vectors of real and fake voice, and use support vector machine to classify real and fake voice in the back-end to reduce the negative effect of the new technique.

Improved Support Vector Machine for Voiceprint Diagnosis of Typical Faults in Power Transformers

The traditional power transformer diagnosis method relies on a lot of experience knowledge and a complex sampling process, which brings great difficulties to the fault diagnosis work. To solve this problem, a fault feature extraction method based on fully adaptive noise set empirical mode decomposition (CEEMDAN) is proposed, and the hunter–prey optimization (HPO) algorithm is used to optimize the support vector machine (SVM) to identify and classify the voice print faults of power transformers. Firstly, the CEEMDAN algorithm is used to decompose the voicemarks into several IMF components. IMF components containing fault information are selected according to the envelope kurtosis index and reconstructed to generate new signal sequences. PCA dimensionality reduction is performed on the reconstructed signal, and the principal components are extracted with a high cumulative contribution rate as input to SVM. Then, the HPO-SVM algorithm is used to classify and identify transformer faults. Apply the proposed method to the diagnosis of typical faults in power transformers. The results show that the accuracy of this method in identifying various fault states of power transformers can reach 98.5%, and it has better classification performance than other similar methods.

Development of Double Channel Transformer Voiceprint Detection System Based on MEMS Digital Silicon Microphone

Aiming for on-line monitoring of typical mechanical faults of transformer, a two-channel MEMS digital silicon wheat voice print detection system is developed, which consists of two channel MEMS digital silicon wheat array, the host of sound acquisition system and the upper computer software with the function of displaying real-time waveform and spectrum analysis. The device can operate reliably in complex environments by adjusting the array spacing, improving the power quality, and adjusting the network protocol. Further, different test environments were built to test the resolution performance of the equipment to the center frequency of the sound source at different distances and angles of the sound source. The results show that the system can realize the resolution of the center frequency of different sound source distance and Angle, and still has a good center frequency resolution performance in the case of environmental noise. At the same time, the system has strong stability, environmental adaptability, easy installation and non-contact detection characteristics, can be used for real-time detection and diagnosis of transformer faults.

A Contrastive- Acoustic Study of Jacqueline Kennedy and the Actress Natalie Portman’s Voice Qualities in the Film” Jacki”.

As performers in a social world, we communicate with other people by sharing information on many different levels. Each utterance includes linguistic information and conveys much information about the speaker’s identity. Variation in voice quality indexes information about the speaker and marks the speaker’s identity as a unique individual. The present study aims to validate the belief that each individual has an inalienable voice print that can’t be imitated. The study verifies that, even the more similar personality between two individuals, or the close position in society the variance is voice quality. The acoustic analysis is performed via analysing the acoustic parameters namely: the fundamental frequency, amplitude, intensity, duration and pitch of Jacqueline Kennedy and the actress Natalie Portman who performed Kennedy’s character in the film “ Jacki”. The results show that Kennedy’s voice quality is breathy; whereas, Portman’s voice quality is falsetto. The study concludes that it is possible to recognise the identity of any individual from the properties of her/his voice quality due to the fact that each person has an unparalleled voice print.

МЕТОДИ ПІДВИЩЕННЯ ЯКОСТІ ПЕРЕТВОРЕННЯ МОВИ НА ТЕКСТ В СИСТЕМАХ БІОМЕТРИЧНОЇ АУТЕНТИФІКАЦІЇ

The article discusses the methods and algorithms of speech-to-text conversion, modern open and commercial systems for creating systems, as well as the use of these technologies in the field of cyber security. It is proposed to create a high-quality speech-to-text conversion system. An analysis of the mathematical algorithms used to reduce the error rate, which makes it possible to create unique voice prints and increase protection against forgery, has been carried out. The structure of modern speech-to-text conversion systems is described. By changing datasets, parameters of hidden Markov models, a high-quality dictionary of phonemes, and the use of language models, there is an opportunity to reduce the percentage of errors in language recognition, as well as the use of a system for multilingualism such as "surzhyk". The mathematical methods of assessing the quality of the system of speech to text (WER), as well as various methods of calculation, which is important for their further improvement and optimization, are considered. The structure of modern systems is considered, namely, signal pre-processing, feature extraction, acoustic modeling, speech modeling, decoding, post-processing. For each of the stages, study vectors have been proposed that can reduce the error rate of the system as a whole. Reducing speech recognition errors and the ability to fake a voice is achieved using various methods: deep neural networks, hidden Markov models, Baum-Welch algorithm, N-gram models, models with attention, creation of a high-quality phonemes dictionary, dataset, and fillers. Speech-to-text conversion technology can be used in biometric authentication systems to detect and analyze the unique features of the user's voice. However, modern speech-to-text conversion systems for Ukrainian, Russian, and "surzhyk" need improvement in acoustic and language units. Scientific works, which are devoted to research and optimization of these systems for biometric authentication, do not fully cover these issues. This became the reason for further research in this direction, so this work aims to create a speech recognition system with a minimum error rate.

Voice Print Recognition Check-in System Based on Resnet

With the development of modern technology and the rise of artificial intelligence, the application scenarios of identity authentication technology are becoming more and more complex, especially the current situation of the spread of the novel coronavirus, the traditional identity authentication technology can no longer meet people's actual needs, and the society urgently needs a secure and convenient identity authentication technology. Voice print recognition technology is an identity recognition method that uses specific feature extraction methods to extract the features that can identify the speaker's identity from the original speech input, and then uses these features to identify the speaker's identity. Aiming at the above problems, this paper proposes a deep learning-based voicing recognition algorithm, which is based on the theoretical knowledge of deep learning. The research work includes the following aspects: providing convolutional network to extract features; The "speech feature template library" is established by massive training. Research on matching and recognition algorithm; Research on check-in system based on voice print recognition. Based on this algorithm, this paper designs and implements a voiceprint recognition check-in system with clear interactive interface. The system has the functions of adding members, refreshing the voice library, checking in with the voice print, clearing the results and so on. The system interface will display the data of the voice print library and the historical check-in record, as well as the current recognition results, accuracy and check-in time. The average recognition rate of the system is about 95%, which can meet the requirement of practical application.

Enhancing Voice Print using Computer Vision through Optical Character Recognition with Subscriber Identity Module

Enhancing Voice Print using Computer Vision through Optical Character Recognition with Subscriber Identity Module

G-Cocktail: An Algorithm to Address Cocktail Party Problem of Gujarati Language Using Cat Boost

This paper is an attempt to address to the problem of native language in a mixed voice environment. G- Cocktail would aid these applications in identifying commands given in Gujarati, even from a mixed voice stream. There are two phases of G-cocktail in the first phase, it creates features after filtering the voices and in the second it trains and classifies the dataset. This trained dataset helps in recognizing the new voice signal. The challenge in training a native language is the availability of a small dataset. A single-word input is used in model and phrase benchmark dataset from Microsoft and the Linguistic Data Consortium for Indian Languages (LDC-IL). To overcome the over fitting problem due to smaller dataset we used CatBoost algorithm. And fine-tuned the classification model to avoid the over fitting issue. Mel-frequency cepstral coefficients (MFCCs) are coefficients that collectively make up an MFC. They are derived from a type of cepstral representation of the audio clip (a nonlinear "spectrum-of-a-spectrum"). MFCC is good for human voices but noises in the sound makes it less productive. To avoid this shortcoming of MFCC, first filtered the voices are used and then calculated the MFCC. The most relevant features are retained to make it more robust. With MFCC features, the pitch of the voices is also added, as pitch could vary with regional changes, mood of the person, age, and knowledge of the language to the speaker. A voice print of the whole sound files is constructed and fed it as features to the classification model. For training and testing 70% and 30% ratio is used in algorithms like K-means, Naïve Bayes, and Light GBM. Proposed model is compared with given data set and results proved that G-cocktail using XBoost performed better than the others under the given scenario in all parameters.

Backdoor Defence for Voice Print Recognition Model Based on Speech Enhancement and Weight Pruning

Voice print recognition is one of the most mature biometric authentication technologies, and the application of deep neural networks (DNNs) has led to a significant improvement in the accuracy of voice print recognition. However, DNN models can be attacked by backdoor attackers, which poses a serious threat to the security of model for voice print recognition. In this paper, a method is proposed for backdoor defence of voice print recognition model based on speech enhancement and weight pruning. Firstly, input samples are perturbed by superimposing various speech patterns, and the backdoor samples are determined based on the randomness (entropy value) of the prediction classes with perturbed inputs from a given deployment model (malicious or benign). Secondly, the backdoor samples are fed into a network (Deep Complex Convolution Recurrent Network) dedicated for speech enhancement, with which the backdoor samples can be denoised by removing the backdoor noise. Finally, the model is pruned using an automatic progressive weight pruning algorithm, which can avoid the accuracy degradation caused by neurons pruning. Experimental results on the AISHELL speech dataset show that the method not only reduces the success rate of backdoor attacks, but also greatly realizes the purification of the backdoor samples.

Optimistic locating method of transformer voiceprint based on beam and delay algorithm

In view of the on-line inspection mode of transformer voiceprint information during operation, an optimistic location method is proposed to assist the acoustic print diagnosis for fault identification, location and further evaluation work. Firstly, according to the directional location algorithm of antenna array (Beamforming), traditional directional location is extended to general spatial location. Then on the basis of TDOA(Time Difference Of Arrival) algorithm of time domain signal characteristic analysis, integrated weighted processing and data centralization, on the one hand, reduce the edge value, on the other hand, correct the trend of bad data, and improve the existing cross-correlation operation. The simulation results show that the optimized beam-based and weighted TDOA algorithm has excellent performance in 2D localization of transformer voiceprint. These research results can provide a basis for transformer online monitoring and voice print diagnosis.

Research on End-to-end Voiceprint Recognition Model Based on Convolutional Neural Network

Speech signal is a time-varying signal, which is greatly affected by individual and environment. In order to improve the end-to-end voice print recognition rate, it is necessary to preprocess the original speech signal to some extent. An end-to-end voiceprint recognition algorithm based on convolutional neural network is proposed. In this algorithm, the convolution and down-sampling of convolutional neural network are used to preprocess the speech signals in end-to-end voiceprint recognition. The one-dimensional and two-dimensional convolution operations were established to extract the characteristic parameters of Meier frequency cepstrum coefficient from the preprocessed signals, and the classical universal background model was used to model the recognition model of voice print. In this study, the principle of end-to-end voiceprint recognition was firstly analyzed, and the process of end-to-end voice print recognition, end-to-end voice print recognition features and Res-FD-CNN network structure were studied. Then the convolutional neural network recognition model was constructed, and the data were preprocessed to form the convolutional layer in frequency domain and the algorithm was tested.

Imperfect Sound Forever

What is phonography? In this essay, Christopher DeLaurenti, a phonographer with three decades of experience, maps an axiomatic 13-lesson pedagogy through an abbreviated history of field recording, from Jesse Walter Fewkes in 1890 to Tony Schwartz in the early 1960s. This paper surveys various meanings and uses of the term phonography from a text published in 1701 to the formation in 2000 of the phonography listserv, an online community of makers of field recordings. The author, himself an early member of the phonography listserv, discusses three traits to define phonography as a community in the early 2000s: inexpensive recording equipment; a community of knowledge; and the “easy fidelity” made possible by portable and lightweight Digital Audio Tape (DAT) and MiniDisc (MD) recorders. The author contrasts the traits of phonography with elements of soundscape composition as articulated by Barry Truax, Hildegard Westerkamp, and Andra McCartney. The paper concludes by proposing possible elements of post-phonography, including remote control recording, the possibility of voice print identification, and the generation of unimagined data.

Building a trusted guardian for our biometric identities

These days, unlocking a phone with a quick glance or the touch of a finger – rather than a passcode – has become so popular that we almost forget that we are using our biometric information to log in. The same can be said for authenticating payments or signing up to the latest challenger bank. In return for convenience, and to prove we are who we say we are, most of us now happily allow commercial companies and other organisations to process our biometric facial, fingerprint, iris or voice print data.

Comparative Study of Speaker Recognition Techniques in IoT Devices for Text Independent Negative Recognition

Speaker recognition (SR) or identification is the subset of broad area of Pattern recognition. Given the features of the voice print, the recognition system identifies the speaker from the knowledge of the speaker models stored in the database. In today’s world when many of our works are done through voice, recognition of the speaker is necessary.Recently, SR has also gained importance in Internet of Things (IoT) like setting up of smart environments for home, industries or educational and commercial applications. The race for high accuracy needs making the devices used in these smart environments as close to human hearing capacity as possible. Speaker identification is mostly used to establish negative recognition. Negative recognition is when thesystem decides whether a person is who he disagrees to be thus preventing a person from exploiting multiple identities. Only biometrics will be suitable to establish such identification. The feature extraction of voice sample along with comparative analysis of its methods is of fundamental interest in this paper. We try to compare the performance of features which are used in state of art speaker recognition models and analyse variants of Mel frequency cepstrum coefficients (MFCC) predominantly used in feature extraction which can be further incorporated and used in various smart devices.

Development of methods for identification of information­controlling signals of unmanned aircraft complex operator

Methods for verifying and identifying the operator by the features of the formation of biometric features of a speech signal in control systems of unmanned aerial systems are proposed. A method has been developed for the effective width of the spectrum of a speech signal, which allows identification and verification of the operator of an unmanned aerial vehicle based on an analysis of the informative components of voice prints under conditions of a high level of interference of various origins. A method has been developed for the highest informational weight of the fundamental tone, which is based on the use of the most informative components of the spectral representation of the prints of a speech signal. The first method allows to identify the operator of an unmanned aerial vehicle by the informative components of the spectral representation of the fingerprint of a speech signal under conditions of a high level of interference. High indicators, which are achieved by using this method, are obtained due to the uniqueness of the selected feature space, which retain their characteristics even with a fairly high level of interference. The second method provides speaker identification of an unmanned aerial vehicle by a specific space of unique voice features. The frequencies of the fundamental tone and overtones were chosen as the basic features. Such an approach to solving the identification problem provides a high probability of determining the operator with the existing rather high level of interference and reduces the processing time of information in comparison with the effective spectrum width method. The creation of control methods and models for unmanned aerial systems provides an increase in the level of noise immunity and safety of control systems from interventions by an unauthorized operator. Using operator identification methods allows to create a system for restricting access to the aircraft control process and thereby ensure the continuity of the operation of the information management system for unmanned aerial systems

Research on Voice Print Recognition Algorithm based on Wavelet-GA-BP Neural Network

Aiming at the fact that common pattern recognition algorithms such as LPCC and MFCC have flaws in voice-print recognition, the paper put forward a new algorithm, which employed wavelet analysis, BP neural network and niche genetic algorithm. Firstly, the algorithm extracted time-domain and frequency-domain characteristic variation of voice signal by wavelet transform, secondly, trained the neural network niche genetic algorithm, which solved the problem of local minimum value caused by normal multi-layers neural network, lastly, took the wavelet transform variations as the training data of the optimization neural network. Simulation experiment was carried out. The results indicate the algorithm in this paper is superior to current recognition algorithms, it has fast recognition velocity, high recognition rate, low fault rate and strong robustness with different voice-makers, besides, it can automatically correct errors.