Speaker Identification Based on Incremental Learning Neural Network

This paper investigates an approach for speaker identification in a remote access system based on coded speech signals. The aim of using the coding process is to decrease the amount of transmitted data via the channel. In this proposed system, the speech signal is coded by two different coding techniques. It can be coded either by linear predictive coding or compressive sensing. The coded speech signal is transmitted into the receiver via the wireless communication channel. At the receiver, the received signal is decoded, and then speaker identification system is applied on the decoded signal. During the transmission process, the channel errors affect on the transmitted signal, so they should be taken into account. The speaker identification process is used to achieve the security needed for the remote access system. In speaker identification system, the feature vectors are captured from different discrete transforms such as discrete wavelet transform, discrete cosine transform, and discrete sine transform, besides the time domain. The recognition rate for all transforms is computed to evaluate the effect of coded signals on the performance of the speaker identification system. The results proved that the discrete cosine transform and discrete wavelet transform are the best. In addition to the proposed system gives close recognition results to those obtained from real speech signals revealing a simple degradation effect due to the speech coding.

Speech processing is one of the required fields in digital signal processing that helps in processing the speech signals. The speech process is utilized in different fields such as emotion recognition, virtual assistants, voice identification, etc. Among the various applications, emotion recognition is one of the critical areas because it is used to recognize people’s exact emotions and eliminate physiological issues. Several researchers utilize signal processing and machine learning techniques together to find the exact human emotions. However, they fail to attain their feelings with less computational complexity and high accuracy. This paper introduces the intelligent computational technique called cat swarm optimized spiking neural network (CSSPNN). Initially, the emotional speech signal is collected from the Toronto emotional speech set (TESS) dataset, which is then processed by applying a wavelet approach to extract the features. The derived features are further examined using the defined classifier CSSPNN, which recognizes human emotions due to the effective training and learning process. Finally, the proficiency of the system is determined using experimental results and discussions. The proposed system recognizes the speech emotions up to 99.3% accuracy compared to recurrent neural networks (RNNs), deep neural networks (DNNs) and deep shallow neural networks (DSNNs).

This paper presents a text independent speaker identification system using multi-band features with artificial neural network. Linear predictive cepstrum coefficients (LPCCs) computed from sub-band signals with higher order statistics (HOS) are employed as the main features to represent the speaker characteristics. The multi-band representation of the speech signal is implemented by empirical mode decomposition (EMD). Dominant feature vectors are derived by applying principal component analysis (PCA) on LPCC space computed from the speech signal. The experimental results show that the proposed system improves the speaker identification performance. The efficiency is also compared for different features with noisy speech signals.

Context. The questions of adapting the convolution neural network classifier use in automatic speaker recognition system of critical use(ASRSCU) are considered. The research object is the individual features of the human speech process.Objective. Development of means for separating individual features from the speaker’s speech signal, increasing their informativeness asa result of the factor analysis, their visual representation for the use of the convolution neural network classifier, and optimizing itsarchitecture for the needs of ASRSCU.Method. Measures are proposed to optimize the speaker recognition procedure of the ASRSCU, for which the optimal way of informativefeatures representation and the method of increasing their informativeness are theoretically justified, the topology and measures for increasingof the speaker recognition process efficiency are justified. In particular, it is justified the use of power normalized cepstral coefficients (PNCC)for the description of phonograms recorded in noisy environment conditions. We propose to use Gabor filters to represent information thatwill be analyzed by a convolution neural network, an optimal method of factor analysis (a sparse main components analyzing method) toreduce of the features vector length while preserving its informativeness, an improved topology of the convolution neural network in whichthe Gabor filters are integrated in to the convolution layer, which allows them to optimize their parameters during the neural network trainingprocess, and in a fully connected layer a deep neural network with a bottleneck layer is used, whose weights after training are uses as inputs forthe GMM/HMM control classifier.Results. Methods of representation and optimization of the speaker’s individual features, methods for their visual presentation andimprovement of the topology of a convolution neural network for making speaker recognition on their basis.Conclusions. The obtained theoretical results have found empirical confirmation. In particular, the stability of an improved convolutionneural network to the noisy input phonograms proved to be higher than the results of an ordinary convolution neural network and a deep neuralnetwork. With an SNR increase up to 10 dB, the GMM/HMM classifier is more efficient than the neural network, which can be explained by the efficiency of the used UBM models, but it is much more resource-intensive. Also, the parameters of the Gabor filter bank frames thatprovide the most variable individual features from the speech signal for speaker recognition are determined empirically.

This paper examines the architecture of back-propagation neural networks used as approximators by addressing the interrelationship between the number of training pairs and the number of input, output, and hidden layer nodes required for a good approximation. It concentrates on nets with an input layer, one hidden layer, and one output layer. It shows that many of the currently proposed schemes for selecting network architecture for such nets are deficient. It demonstrates in numerous examples that overdetermined neural networks tend to give good approximations over a region of interest, while underdetermined networks give approximations which can satisfy the training pairs but may give poor approximations over that region of interest. A scheme is presented that adjusts the number of hidden layer nodes in a neural network so as to give an overdetermined approximation. The advantages and disadvantages of using multiple output nodes are discussed. Guidelines for selecting the number of output nodes are presented.

Acoustical communication is one of the fundamental prerequisites for the existence of human society. Textual language has become extremely important in modern life, but speech has dimensions of richness that text cannot approximate. From speech alone, fairly accurate guesses can be made as to whether the speaker is male or female, adult or child. In addition, experts can extract from speech information regarding e.g. the speaker’s state of mind. As computer power increased and knowledge about speech signals improved, research of speech processing became aimed at automated systems for many purposes. Speaker recognition is the complement of speech recognition. Both techniques use similar methods of speech signal processing. In automatic speech recognition, the speech processing approach tries to extract linguistic information from the speech signal to the exclusion of personal information. Conversely, speaker recognition is focused on the characteristics unique to the individual, disregarding the current word spoken. The uniqueness of an individual’s voice is a consequence of both the physical features of the person vocal tract and the person mental ability to control the muscles in the vocal tract. An ideal speaker recognition system would use only physical features to characterize speakers, since these features cannot be easily changed. However, it is obvious that the physical features as vocal tract dimensions of an unknown speaker cannot be simply measured. Thus, numerical values for physical features or parameters would have to be derived from digital signal processing parameters extracted from the speech signal. Suppose that vocal tracts could be effectively represented by 10 independent physical features, with each feature taking on one of 10 discrete values. In this case, 1010 individuals in the population (i.e., 10 billion) could be distinguished whereas today’s world population amounts to approximately 7 billion individuals. People can reliably identify familiar voices. About 2-3 seconds of speech is sufficient to identify a voice, although performance decreases for unfamiliar voices. One review of human speaker recognition (Lancker et al., 1985) notes that many studies of 8-10 speakers (work colleagues) yield in excess of 97% accuracy if a sentence or more of the test speech is heard. Performance falls to about 54% when duration is shorter than 1 second and/or distorted e.g., severely highpass or lowpass filtered. Performance also falls significantly if training and test utterances are processed through different transmission systems. A study

При тестировании на специализированных базах данных наиболее совершенных систем идентификации диктора их минимальная эффективность, оцениваемая величиной вероятности ошибки в точке пересечения кривых ошибок, составляет лишь несколько процентов. Однако известно множество факторов, влияющих на вариативность характеристик голоса диктора, каждый из которых имеет свое, отличное от других, влияние на результаты идентификации диктора по характеристикам голоса. Сложность создания и тестирования систем идентификации диктора заключается в необходимости количественной формализации ряда конкретных факторов, влияющих на характеристики его голоса. Рассмотрен предложенный метод учета множества факторов, влияющих на параметры характеристик голоса диктора, обеспечивающий принципиальную возможность косвенного учета их практически неограниченного количества. Согласно этому методу из речевых сигналов выделяются «атомарные» структуры, зависящие от совокупности основных факторов, влияющих на процесс идентификации диктора. По такому методу все существенные факторы, влияющие на характеристики голоса, будут косвенно учитываться на уровне этих структур. Экспертные решения принимаются по комбинаторной совокупности огромного числа этих «атомарных» структур. Под «атомарными» структурами речи понимаются спектры любых фрагментов гласных звуков, выделяемых во временном окне продолжительностью 20 мс. "Атомарные" структуры выделяются в автоматическом режиме. Предлагаемый метод обеспечивает рациональный учет богатофакторности влияния различных параметров, поскольку на спектры этих структур влияют все основные факторы, характеризующие индивидуальность голоса конкретного диктора. Решение об идентичности голосов дикторов, записанных на разных фонограммах, осуществляется на основе комбинаторики «атомарных» спектров гласных звуков в обоих фонограммах. Метод показал высокую эффективность при экспертизе фонограмм малой продолжительности.

Traditional neural network-based speaker identification (SI) studies employ a combination of acoustic features extracted from sequential sounds to present the speakers’ voice biometrics in which several sound segments before and after the current segment are stacked and fed to the network. Although this method is particularly important for speech recognition tasks where words are constructed from sequential sound segments, and successful recognition of words depends on the previous phonetic sequences, SI systems should be able to operate based on the distinctive speaker features available in an individual sound segment and identify the speaker regardless of the previously uttered sounds. This paper investigates this hypothesis by proposing a novel text-independent SI model trained at sound level. In order to achieve this, the investigation was conducted by first studying the best distinguishable configuration of coefficients in a single acoustic segment, then to identify the best frame length to overlapping ratio, and finally measuring the reliability of conducting SI using only a single sound segment. Overall more than one hundred SI systems were trained and evaluated, in which results indicate that performing SI using a single acoustic sound frame decreases the complexity of SI and facilitates it since the classifier requires to learn fewer number of acoustic features in compare to the traditional stacked-based approaches.

This paper presents a novel speech enhancement approach by combining Fourier series expansion and spectral subtraction. This approach is implemented in speaker identification systems where degraded speech could result in high false speaker identifications. A Fourier series is estimated for the noisy speech signals, and then spectral subtraction is used to reduce the amount of noise in order to enhance quality of the speech signals before the speaker identification process. Experimental results presented to compare between the proposed approach and the traditional methods demonstrate the ability of the proposed approach to both enhance speech quality and improve speaker recognition rates.

This work presents an innovative deep neural network-based speaker verification (SV) model. SV is an authentication process that confirms whether a speaker is authorized. Text Independent Speaker Verification identifies who is speaking, not what is being said. India’s diverse linguistic and cultural landscape introduces challenges such as dialect and accent variations, multilingualism, multi-speaker scenarios, and varying quality of voice data recorded. Literature hints that the EER of the speaker verification model shoots up if tested with multi-language and multi-channel data; we have used the IITG-MV dataset, which has these variations captured specially for Indian conditions. Currently, convolutional neural networks (CNNs) have shown encouraging results in speaker verification. Our work presents CNN combined with the LSTMs, where CNN is a feed-forward neural network, and LSTMs are a recurrent neural network. The pipeline extracts the acoustic characteristics from the input speech signal using MFCC and fed into models based on deep neural networks like VGG-16. However, the raw speech signal is fed into the SincNet model. The CNN-LSTMs model demonstrates significantly superior accuracy to SincNet and VGG16, making it the most effective approach for text-independent speaker verification in diverse Indian languages and channels. Despite using a basic deep neural network, the results demonstrate the efficiency of our proposed architecture. CNN-LSTM’s state-of-the-art performance is closer to that of the ECAPA-TDNN model.

Institutions have been adopting work/study-from-home programs since the pandemic began. They primarily utilise Voice over Internet Protocol (VoIP) software to perform online meetings. This research introduces a new method to enhance VoIP calls experience using deep learning. In this paper, integration between two existing techniques, Speaker Separation and Speaker Identification (SSI), is performed using deep learning methods with effective results as introduced by state-of-the-art research. This integration is applied to VoIP system application. The voice signal is introduced to the speaker separation and identification system to be separated; then, the “main speaker voice” is identified and verified rather than any other human or non-human voices around the main speaker. Then, only this main speaker voice is sent over IP to continue the call process. Currently, the online call system depends on noise cancellation and call quality enhancement. However, this does not address multiple human voices over the call. Filters used in the call process only remove the noise and the interference (de-noising speech) from the speech signal. The presented system is tested with up to four mixed human voices. This system separates only the main speaker voice and processes it prior to the transmission over VoIP call. This paper illustrates the algorithm technologies integration using DNN, and voice signal processing advantages and challenges, in addition to the importance of computing power for real-time applications.

We present a feature enhancement method that uses neural networks (NNs) to map the reverberant feature in a log-melspectral domain to its corresponding anechoic feature. The mapping is done by cascade NNs trained using Cascade2 algorithm with an implementation of segment-based normalization. Experiments using speaker identification (SID) and automatic speech recognition (ASR) systems were conducted to evaluate the method. The experiments of SID system was conducted by using our own simulated and real reverberant datasets, while the CENSREC-4 evaluation framework was used as the evaluation for the ASR system. The proposed method could remarkably improve the performance of both systems by using limited stereo data and low speaker-variant data as the training data. From the evaluation using SID, we reached 26.0% and 34.8% of error rate reduction (ERR) relative to the baseline by using simulated and real data, respectively, by using only one pair of utterances for matched condition cases. Then, by using combined dataset containing 15 pairs of utterances by one speaker from three positions in a room, we could reach 93.7% of average identification rate (three known and two unknown positions), which was 42.2% of ERR relative to the use of cepstral mean normalization (CMN). From the evaluation using ASR, by using 40 pairs of utterances as the NN training data, we could reach 78.4% of ERR relative to the baseline by using simulated utterances by five speakers. Moreover, we could reach 75.4% and 71.6% of ERR relative to the baseline by using real utterances by five speakers and one speaker, respectively.

Determining the best model for the neural network architecture and how to optimize the architecture with the metaheuristic Protis Approach is a subject of the study. A comprehensive investigation and utilization of metaheuristic methods are necessary. These methods aim to solve problems and adapt from the lifestyle of the amoeba protis. In this study, the proposed method modifies the life cycle of the amoeba, which consists of four phases: prophase, metaphase, anaphase, and telophase. These four phases are modified in the neural network architecture to optimize the appropriate number of hidden layers and produce an efficient architecture model. The results show that the protis approach optimized the neural network architecture, especially in generating hidden layers to improve the neural network model. Distinctive features of the results obtained are that the average range of degenerate neurons in the hidden layer is 0 to 35 neurons in each layer. The standard number of neurons makes it possible to solve the problem of determining the best model on the neural network architecture. The protis algorithm embedded in the protis recurrent neural network for categorical data measurements produces an average RMSE value, representing the difference between actual measurements and predictions, equal to 0.066. Consequently, the developed model surpasses the current classical neural network model in terms of performance. Regarding accuracy, the protis algorithm embedded in the neural network for categorical and time series data achieves an average precision of 0.952 and a recall of 0.950. The protis convolutional neural network achieves an accuracy of 95.9 %. Therefore, from the three tested datasets, the protis convolutional neural network exhibits the highest accuracy value

Existing extreme learning algorithm have not taken into account four issues: 1) complexity; 2) uncertainty; 3) concept drift; and 4) high dimensionality. A novel incremental type-2 meta-cognitive extreme learning machine (ELM) called evolving type-2 ELM (eT2ELM) is proposed to cope with the four issues in this paper. The eT2ELM presents three main pillars of human meta-cognition: 1) what-to-learn; 2) how-to-learn; and 3) when-to-learn. The what-to-learn component selects important training samples for model updates by virtue of the online certainty-based active learning method, which renders eT2ELM as a semi-supervised classifier. The how-to-learn element develops a synergy between extreme learning theory and the evolving concept, whereby the hidden nodes can be generated and pruned automatically from data streams with no tuning of hidden nodes. The when-to-learn constituent makes use of the standard sample reserved strategy. A generalized interval type-2 fuzzy neural network is also put forward as a cognitive component, in which a hidden node is built upon the interval type-2 multivariate Gaussian function while exploiting a subset of Chebyshev series in the output node. The efficacy of the proposed eT2ELM is numerically validated in 12 data streams containing various concept drifts. The numerical results are confirmed by thorough statistical tests, where the eT2ELM demonstrates the most encouraging numerical results in delivering reliable prediction, while sustaining low complexity.

Abstract— Audio data carries rich information in the form of speech, music, and environmental sounds, but its raw waveform is often complex and high-dimensional, making direct analysis difficult. Neural network architectures have emerged as powerful tools for extracting meaningful representations from audio signals, enabling efficient analysis and interpretation. Recurrent Neural Networks (RNNs), and Transformer-based models— for learning robust and discriminative features from audio. By automatically capturing temporal, spectral, and contextual patterns, these architectures significantly improve performance in tasks such as speech recognition, speaker identification, music classification, and environmental sound detection. The findings highlight the potential of neural networks to replace traditional handcrafted features, thereby advancing the development of scalable, accurate, and realtime audio processing applications. The rapid growth of audio data across domains such as speech, music, healthcare, and environmental monitoring has created a strong need for effective methods to extract meaningful representations from complex audio signals. Traditional approaches rely on handcrafted features like MFCCs and spectrogram descriptors, which often fail to capture the full temporal and spectral dynamics present in raw audio. Keywords: Neural Networks, Deep Learning, Audio Representation Learning, Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Transformer Models, Feature Extraction, Speech Recognition, Speaker Identification, Sound Event Detection, Spectrogram Analysis, Audio Signal Processing