Cost-Sensitive Learning for Emotion Robust Speaker Recognition

Speaker recognition has made great progress under the laboratory environment, but in real life the performance of speaker recognition system is affected by various factors including environmental noise. This paper studies the performance of speaker recognition system in noisy environment and presents Speaker recognition system using modified Mel-Frequency Cepstral Coefficients (MFCC) technique based on different classifiers likes Euclidean distance, Back-Propagation Neural Network (BPNN), Self Organizing Map (SOM). Modified Mel-Frequency Cepstral Coefficients (MFCC) technique includes Blackman windowing instead of hamming window. This paper presents comparative plots of different classifiers based on modified Mel-Frequency Cepstral Coefficients (MFCC) technique. Speaker recognition system based on SOM Neural Network classifier provides better recognition rate compare to BPNN and Euclidean Distance based systems.

This paper investigates the influence of GSM speech coding on the performance of a text independent Speaker Recognition System (SRS). The SRS developed perform recognition on reconstructed speech waveform from the coded parameters using Gaussian Mixture Models (GMM) technique. The performance evaluation due to the use of the GSM speech coding namely the GSMEFR (Global System Mobile Enhanced Full Rate) codec was conducted, using three transcoded databases, obtained by passing the local ARADIGIT database through the GSM coder/decoder. The recognition evaluation was also conducted using original ARADIGIT sampled at 16 KHz and its 8 KHz downsampled version. The ARADIGIT database consists of 60 speakers (31 male speakers and 29 female speakers) pronouncing the ten Arabic digits five time each. Several experiments were conducted in order to evaluate the degradation introduced by different aspects of the simulated coder.

Combination of deep learning and I-vector can significantly improve the performance of speaker recognition system, however, how to optimize the traditional feature parameters and how to model the speaker recognition system through deep learning are two most important research topics. This paper explores the system recognition performance from the types of input and neural network, and researches the optimal feature parameters and the most appropriate neural network structure of a speaker recognition system. Furthermore, the existing speaker recognition algorithms based on deep learning (Deep Neutral Network (DNN), Convolutional Neural Network (CNN)) have been analyzed and several improved models based on deep learning networks have been implemented and compared. Experiment shows that the network model after combination has a higher recognition rate in speaker recognition than the traditional system model.

Voice biometrics are being implemented globally in large scale applications such as remote banking, government e-services, transportation and building security access, autonomous vehicles, and healthcare. They have been integrated in numerous apps, often coupled with face biometrics and artificial intelligence methods. Voice biometrics products and solutions must meet three key re-quirements for the success in their deployment: they must be highly trustable regarding privacy protection; easy to use and always be available. This edited book presents the state of the art in voice biometrics research and technologies including implementation and deployment challenges in terms of interoperability, scalability and performance, and security. The team of editors and chapter authors combine a wealth of expertise from academia and the industry. Topics covered include the fundamentals of voice biometrics; design of countermeasures for replay attack; attacker's perspective for voice biometrics; voice biometrics; speaker de-identification; performance evaluation of voice biometrics solutions; standardization of voice biometrics technology; industry perspectives; joining forces of voice and facial biometrics; and future trends and challenges in voice biometrics. Providing comprehensive coverage of the field of voice biometrics, this authoritative volume will be of great interest to researchers, scientists, engineers, practitioners and advanced students involved in the fields of security, biometrics, forensic sciences, human computer interaction, speech processing, acoustics, multimedia, pattern recognition, and privacy-preserving, digital signal processing and speech technologies. It will also be of interest to researchers and professionals working in law and criminology.

Technologies of virtual and augmented reality are gaining popularity. Virtual reality is used in many areas, including beginning to be used in the field of information and computer security. In addition, virtual reality interfaces are also susceptible to attacks. However, there are still few works on research in this area. This article provides an overview of existing solutions to computer security problems using virtual and augmented reality interfaces, as well as an overview and classification of the identified threats for these interfaces themselves.

Advanced encryption and decryption strategies are of great significance for information protection and data security. Visual optical information encryption and decryption technology plays an important role in the field of information security. However, the current optical information encryption technologies have shortcomings such as the need for external decryption equipment, the inability to read out repeatedly, and information leakage, which hinder their practical application. By combining the excellent thermal response characteristics of the MXene-isocyanate propyl triethoxy silane (IPTS)/polyethylene (PE) bilayer and the structural color generated from the laser fabricated biomimetic structural color surface, an approach of encrypt, decrypt, and transmit information has been proposed. The microgroove-induced structural color is attached to the MXene-IPTS/PE bilayer to form a colored soft actuator (CSA) to realize information encryption and decryption, and information transmission. Benefiting from the unique photon-thermal response of the bilayer actuator and the precise spectrum response of the microgroove-induced structural color, the information encryption and decryption system has the advantages of being simple and reliable, which has the potential application in the field of optical information security.

Speaker and language recognition in noisy and degraded channel conditions continue to be a challenging problem mainly due to the mismatch between clean training and noisy test conditions. In the presence of noise, the most reliable portions of the signal are the high energy regions which can be used for robust feature extraction. In this paper, we propose a front end processing scheme based on autoregressive (AR) models that represent the high energy regions with good accuracy followed by a modulation filtering process. The AR model of the spectrogram is derived using two separable time and frequency AR transforms. The first AR model (temporal AR model) of the sub-band Hilbert envelopes is derived using frequency domain linear prediction (FDLP). This is followed by a spectral AR model applied on the FDLP envelopes. The output 2-D AR model represents a low-pass modulation filtered spectrogram of the speech signal. The band-pass modulation filtered spectrograms can further be derived by dividing two AR models with different model orders (cut-off frequencies). The modulation filtered spectrograms are converted to cepstral coefficients and are used for a speaker recognition task in noisy and reverberant conditions. Various speaker recognition experiments are performed with clean and noisy versions of the NIST-2010 speaker recognition evaluation (SRE) database using the state-of-the-art speaker recognition system. In these experiments, the proposed front-end analysis provides substantial improvements (relative improvements of up to 25%) compared to baseline techniques. Furthermore, we also illustrate the generalizability of the proposed methods using language identification (LID) experiments on highly degraded high-frequency (HF) radio channels and speech recognition experiments on noisy data.

An urgent problem in the field of information security – the provision of human resources in the field of information security in the Russian Federation is presented. One of the factors influencing this is the lack of the possibility of determining at the initial stage the optimal choice of the future direction of activity of graduates of higher educational institutions with specialization in this field. To solve this problem, it is proposed to use methods of data mining, machine learning and algorithms based on artificial neural networks. For the successful application of data mining methods, several factors should be taken into account: the data set must be large enough. Secondly; the data must be structured and analyzed to extract the most important information; the data analysis methods must be adequate and accurate in order to determine the purpose of the graduate with the highest accuracy. Taking into account these factors, the article suggests a way to form the initial data set and prepare it using statistical methods to predict the success of the professional activity of an information security specialist. The results of this work can be useful for higher educational institutions, specialists in the field of information security, heads of personnel departments who are interested in optimizing the process of selecting qualified employees in this field.

In real-life applications, the performance of speaker recognition systems always degrades when there is a mismatch between training and evaluation data. Many domain adaptation methods have been successfully used for eliminating the domain mismatches in speaker recognition. However, usually both training and evaluation data themselves can be composed of several subsets. These inner variances of each dataset can also be considered as different domains. Different distributed subsets in source or target domain dataset can also cause multi-domain mismatches, which are influential to speaker recognition performance. In this study, we propose to use adversarial training for multi-domain speaker recognition to solve the domain mismatch and the dataset variance problems. By adopting the proposed method, we are able to obtain both multi-domain-invariant and speaker-discriminative speech representations for speaker recognition. Experimental results on DAC13 dataset indicate that the proposed method is not only effective to solve the multi-domain mismatch problem, but also outperforms the compared unsupervised domain adaptation methods.

Speaker Recognition is one of the principle problems in Speech processing. The performance of speaker recognition systems can be improved by carefully choosing and calculating suitable features, which is an arduous task. Therefore, the learning based approach has been found to be simpler, more general and with the rapid growth in Artificial Intelligence, more accurate. This paper is a comparative study of the performance of different neural networks in speaker recognition. The focus of this work is to find which of these learning algorithms is more accurate, less complex, and more generic when it comes to speaker recognition. A database of 5000 utterances, 100 for each of the 50 different speakers, in both clean and noisy environment, with varying levels of noise was used. The MFCC (Mel Frequency Cepstral Coefficients) of these utterances were used as features to train and evaluate the neural networks. Accuracy of all neural networks was expectedly very high (>90%) for clean data, large variations coming in with introduction and change in the level of noise. RBFNN has been shown to consistently perform well under all conditions. DNN was the other consistent performer and has the potential to outperform other techniques, if trained on more data.

Identity recognition using the specific biometrical characteristics of a person has recently become a popular technique.Compared with image-sensor-data-based face and fingerprint recognition, speaker recognition using the acoustic characteristics of the uttered voices obtained from a speaking person is an additional alternative.In certain cases of dark environments or dirty fingers, acoustics-based speaker recognition will be an alternative method for accomplishing identity recognition with satisfactory recognition accuracy.Speaker recognition in practical application scenarios will inevitably encounter the problem of acoustic speech mixed with background noises.Utterances with undesired background noises of specific environments cannot be finely matched with the preestablished speaker models, thus causing inaccurate identity recognition results.To tackle this issue, we present a deep-learning-based method for speaker recognition in a noisy environment, which is a hybridization of two different types of deep learning calculation model, speech enhancement generative adversarial network (SEGAN) and convolutional neural network (CNN), called hybridized SEGAN-CNN.By removing specific background noise from the substandard utterance with noise using SEGAN and classifying the identities of numerous speaking subjects without noise effects using CNN, the task becomes speaker recognition in a clear environment, in which the robustness of speaker recognition can be effectively maintained.The results of experiments using a voice command phrase mixed with motor operation noise for robot navigation control in a simulated factory environment demonstrate the effectiveness of the proposed speaker recognition method.

We note that the effective and active struggle of European countries with information terrorism takes place within the framework of the European Union, all its member-countries, as well as neighboring countries and applicants for accession. Therefore, it should be stressed that the EU is pursuing itself a very active policy in the field of information security. At the moment, it brings together highly developed countries that have a tremendous impact on international relations by establishing norms and standards of conduct of states in the political, economic, social, informational and other spheres. The main objective of the EU is to strengthen the European Commission’s dialogue with international organizations and partners on the issue of network security and, in particular, on the growing dependence on electronic networks. Political priorities in the field of information security, defined by the governing bodies of the European Union, are being implemented at the national level by both state authorities and nongovernmental organizations. Analyzing all the above, it can be concluded that within the framework of the EU, information security is considered, first of all, as a state of information networks and systems that provides an adequate level of protection of the integrity, availability, authenticity and confidentiality of information and the appropriate level of counteraction to external negative influences. Priorities of EU policy in the field of information security are the creation and implementation of programs and various technical means of protection of information and communication technologies; development of normative legal acts that establish a list of crimes in the IT sphere and criminal liability; ensuring a high level of public awareness of the risks, threats and ways of protecting their information systems / networks from undesirable effects.

The QUT-NOISE-SRE protocol is designed to mix the large QUT-NOISE database, consisting of over 10 hours of back- ground noise, collected across 10 unique locations covering 5 common noise scenarios, with commonly used speaker recognition datasets such as Switchboard, Mixer and the speaker recognition evaluation (SRE) datasets provided by NIST. By allowing common, clean, speech corpora to be mixed with a wide variety of noise conditions, environmental reverberant responses, and signal-to-noise ratios, this protocol provides a solid basis for the development, evaluation and benchmarking of robust speaker recognition algorithms, and is freely available to download alongside the QUT-NOISE database. In this work, we use the QUT-NOISE-SRE protocol to evaluate a state-of-the-art PLDA i-vector speaker recognition system, demonstrating the importance of designing voice-activity-detection front-ends specifically for speaker recognition, rather than aiming for perfect coherence with the true speech/non-speech boundaries.

Although the field of automatic speaker or speech recognition has been extensively studied over the past decades, the lack of robustness has remained a major challenge. The missing data technique (MDT) is a promising approach. However, its performance depends on the correlation across frequency bands. This paper presents a new reconstruction method for feature enhancement based on the trait. In this paper, the degree of concentration across frequency bands is measured with principal component analysis (PCA). Through theoretical analysis and experimental results, it is found that the correlation of the feature vector extracted from the sub-band (SB) is much stronger than the ones extracted from the full-band (FB). Thus, rather than dealing with the spectral features as a whole, this paper splits full-band into sub-bands and then individually reconstructs spectral features extracted from each SB based on MDT. At the end, those constructed features from all sub-bands will be recombined to yield the conventional mel-frequency cepstral coefficient (MFCC) for recognition experiments. The 2-sub-band reconstruction approach is evaluated in speaker recognition system. The results show that the proposed approach outperforms full-band reconstruction in terms of recognition performance in all noise conditions. Finally, we particularly discuss the optimal selection of frequency division ways for the recognition task. When FB is divided into much more sub-bands, some of the correlations across frequency channels are lost. Consequently, efficient division ways need to be investigated to perform further recognition performance.

ABSTRACT: Speech is one of the most natural and efficient forms of communication among humans, and it is becoming increasingly integral to human-computer interaction in modern applications such as virtual assistants, smart devices, and automated customer service systems. Among the various technologies that leverage speech, speaker recognition stands out as a biometric method that identifies or verifies individuals based on the unique characteristics of their voice. This approach has gained significant traction due to its non-intrusive nature, ease of integration, and applicability across various domains such as security, forensics, and personalised user experiences. In this paper, we introduce a novel speaker recognition system specifically designed to identify speakers based on utterances in the Marathi language, a linguistically rich and widely spoken regional language in India. The system utilises Mel-Frequency Cepstral Coefficients (MFCCS) to extract distinguishing vocal features that closely mimic the human auditory perception. MFCCS are particularly effective in capturing speech’s phonetic and acoustic properties, making them a preferred choice in speech and speaker recognition tasks. Vector quantisation (VQ) is applied to optimise the feature set and reduce computational complexity. VQ compresses the high-dimensional MFCC feature vectors into representative clusters without significantly compromising accuracy, thereby enhancing system efficiency. These features are then processed using hidden Markov Models (HMMS), which are adept at modelling temporal sequences and dynamic variations in speech. HMMS offer a statistical framework that effectively captures the sequential nature of speech patterns, leading to more reliable speaker modelling and recognition.