Speaker Recognition Algorithm Research Articles

A novel speech feature fusion algorithm for text-independent speaker recognition

A novel speech feature fusion algorithm for text-independent speaker recognition

Speaker Gender Recognition Based on Deep Neural Networks and ResNet50

Several speaker recognition algorithms failed to get the best results because of the wildly varying datasets and feature sets for classification. Gender information helps reduce this effort since categorizing the classes based on gender may help lessen the impact of gender variability on the retrieved features. This study attempted to construct a perfect classification model for language-independent gender identification utilizing the Common Voice dataset (Mozilla). Most previous studies are doing manual extracting characteristics and feeding them into a machine learning model for categorization. Deep neural networks (DNN) were the most effective strategy in our research. Nonetheless, the main goal was to take advantage of the wealth of information included in voice data without requiring significant manual intervention. We trained the deep learning network to choose essential information from speech spectrograms for the classification layer, performing gender detection. The pretrained ResNet 50 fine-tuned gender data successfully achieved an accuracy of 98.57% better than the traditional ML approaches and the previous works reported with the same dataset. Furthermore, the model performs well on additional datasets, demonstrating the approach’s generalization capacity.

Research on x-vector speaker recognition algorithm based on Kaldi

Research on x-vector speaker recognition algorithm based on Kaldi

Research on x-vector speaker recognition algorithm based on Kaldi

Research on x-vector speaker recognition algorithm based on Kaldi

A strong hybrid AdaBoost classification algorithm for speaker recognition

A strong hybrid AdaBoost classification algorithm for speaker recognition

Channel Mismatch Speaker Verification Based on Deep Learning and PLDA

At present, speaker recognition technology is developing continuously, but in real world, existing speaker recognition algorithms are generally disturbed by channel factors. With the increase of current data volume, neural network has also become a key technology for speaker recognition. This paper introduction a speaker recognition method based on supervised training and a channel compensation algorithm. The experimental results show that this method combined with channel compensation algorithm can remove the influence of channel noise and has a good speaker recognition effect. The experimental results show that the combination of x-vector and PLDA channel compensation algorithm can obtain better recognition rate and ensure the robustness of the algorithm.

Noise-resistant speaker recognition algorithm based on full-frequency speech features with wavelet packet

At present, most speaker recognition algorithms are performed in a clean environment, and the effect is poor in a noisy environment. In order to improve the accuracy of speaker recognition in a noisy environment, a new feature extraction method, wavelet packet & Gammatone (WPGT) based model, is proposed. In this model, the wavelet packet is used to decompose high-frequency and low-frequency signals and the Gammatone filter bank simulates the human auditory system to process non-linear signals so that more complete speaker voice features are extracted, and finally, the convolutional neural network is used to train the features and complete speaker recognition. Based on the open source speech data sets and the noise fusion data sets, the proposed method is compared with the commonly used voiceprint feature extraction methods MFCC and Gammatone. The experimental results show that, in a noisy environment, WPGT has better anti-noise ability than MFCC and Gammatone. Compared with MFCC and Gammatone, the accuracy of WPGT is improved by 10.63% and 16.91%, respectively.

Dietary Composition Perception Algorithm Using Social Robot Audition for Mandarin Chinese

As the problem of an aging population becomes more and more serious, social robots have an increasingly significant influence on human life. By employing regular question-and-answer conversations or wearable devices, some social robotics products can establish personal health archives. But those robots are unable to collect diet information automatically through robot vision or audition. A healthy diet can reduce a person's risk of developing cancer, diabetes, heart disease, and other age-related diseases. In order to automatically perceive the dietary composition of the elderly by listening to people's chatting, this paper proposed a chat-based automatic dietary composition perception algorithm (DCPA). DCPA uses social robot audition to understand the semantic information and percept dietary composition for Mandarin Chinese. Firstly, based on the Mel-frequency cepstrum coefficient and convolutional neural network, a speaker recognition method is designed to identify speech data. Based on speech segmentation and speaker recognition algorithm, an audio segment classification method is proposed to distinguish different speakers, store their identity information and the sequence of expression in a speech conversation. Secondly, a dietetic lexicon is established, and two kinds of dietary composition semantic understanding algorithms are proposed to understand the eating semantics and sensor dietary composition information. To evaluate the performance of the proposed DCPA algorithm, we implemented the proposed DCPA in our social robot platform. Then we established two categories of test datasets relating to a one-person and a multi-person chat. The test results show that DCPA is capable of understanding users' dietary compositions, with an F1 score of 0.9505, 0.8940 and 0.8768 for one-person talking, a two-person chat and a three-person chat, respectively. DCPA has good robustness for obtaining dietary information.

Fusing MFCC and LPC Features Using 1D Triplet CNN for Speaker Recognition in Severely Degraded Audio Signals

Speaker recognition algorithms are negatively impacted by the quality of the input speech signal. In this work, we approach the problem of speaker recognition from severely degraded audio data by judiciously combining two commonly used features: Mel Frequency Cepstral Coefficients (MFCC) and Linear Predictive Coding (LPC). Our hypothesis rests on the observation that MFCC and LPC capture two distinct aspects of speech, viz., speech perception and speech production. A carefully crafted 1D Triplet Convolutional Neural Network (1D-Triplet-CNN) is used to combine these two features in a novel manner, thereby enhancing the performance of speaker recognition in challenging scenarios. Extensive evaluation on multiple datasets, different types of audio degradations, multi-lingual speech, varying length of audio samples, etc. convey the efficacy of the proposed approach over existing speaker recognition methods, including those based on iVector and xVector.

Genetic Algorithm for Combined Speaker and Speech Recognition using Deep Neural Networks

Huge growth is observed in the speech and speaker recognition field due to many artificial intelligence algorithms being applied. Speech is used to convey messages via the language being spoken, emotions, gender and speaker identity. Many real applications in healthcare are based upon speech and speaker recognition, e.g. a voice-controlled wheelchair helps control the chair. In this paper, we use a genetic algorithm (GA) for combined speaker and speech recognition, relying on optimized Mel Frequency Cepstral Coefficient (MFCC) speech features, and classification is performed using a Deep Neural Network (DNN). In the first phase, feature extraction using MFCC is executed. Then, feature optimization is performed using GA. In the second phase training is conducted using DNN. Evaluation and validation of the proposed work model is done by setting a real environment, and efficiency is calculated on the basis of such parameters as accuracy, precision rate, recall rate, sensitivity, and specificity. Also, this paper presents an evaluation of such feature extraction methods as linear predictive coding coefficient (LPCC), perceptual linear prediction (PLP), mel frequency cepstral coefficients (MFCC) and relative spectra filtering (RASTA), with all of them used for combined speaker and speech recognition systems. A comparison of different methods based on existing techniques for both clean and noisy environments is made as well.

Curriculum Learning Based Approaches for Noise Robust Speaker Recognition

Performance of speaker identification (SID) systems is known to degrade rapidly in the presence of mismatch such as noise and channel degradations. This study introduces a novel class of curriculum learning (CL) based algorithms for noise robust speaker recognition. We introduce CL-based approaches at two stages within a state-of-the-art speaker verification system: at the i-Vector extractor estimation and at the probabilistic linear discriminant (PLDA) back-end. Our proposed CL-based approaches operate by categorizing the available training data into progressively more challenging subsets using a suitable difficulty criterion. Next, the corresponding training algorithms are initialized with a subset that is closest to a clean noise-free set, and progressively moving to subsets that are more challenging for training as the algorithms progress. We evaluate the performance of our proposed approaches on the noisy and severely degraded data from the DARPA RATS SID task, and show consistent and significant improvement across multiple test sets over a baseline SID framework with a standard i-Vector extractor and multisession PLDA-based back-end. We also construct a very challenging evaluation set by adding noise to the NIST SRE 2010 C5 extended condition trials, where our proposed CL-based PLDA is shown to offer significant improvements over a traditional PLDA based back-end.

Speaker Recognition Exploiting D2D Communications Paradigm: Performance Evaluation of Multiple Observations Approaches

The diffusion of Device-to-Device (D2D) communications opens the door to exploit the contributions of multiple Mobile Devices (MDs) to accomplish collaborative tasks. In this paper a speaker recognition algorithm for MDs based on a multiple-observations approach is presented. We propose various fusion and clustering algorithms aimed at efficiently exploiting data coming from MDs. Numerical results show that in many cases our multiple-observation approach is able to significantly improve the accuracy of the considered speaker recognition algorithm.

Real Time Speaker Recognition System using MFCC and Vector Quantization Technique

This paper represents a very strong mathematical algorithm for Automatic Speaker Recognition (ASR) system using MFCC and vector quantization technique in the digital world. MFCC and vector quantization techniques are the most preferable and promising these days so as to support a technological aspect and motivation of the significant progress in the area of voice recognition. Our goal is to develop a real-time speaker recognition system that has been trained for a particular speaker and verifies the speaker. ASR is a type of biometric that uses an individual’s voice for recognition processes. Speaker-vocal discriminative parameters exist in speech signals and due to dissimilar resonances of different speakers speaker recognition system verifies the speaker. These different characteristics can be accomplished by extracting features in vector form like MelFrequency Cepstral Coefficient (MFCCs) from the audio signal. The Vector Quantization (VQ) technique maps vectors from a large vector space to a limited number of regions in the same multidimensional space. LBG (Linde, Buzo and Gray) algorithm is mostly used and preferred for clustering a set of L acoustic vectors into a set of M codebook vectors in speaker recognition.

Speaker identification using vowels features through a combined method of formants, wavelets, and neural network classifiers

This paper proposes a new method for speaker feature extraction based on Formants, Wavelet Entropy and Neural Networks denoted as FWENN. In the first stage, five formants and seven Shannon entropy wavelet packet are extracted from the speakers’ signals as the speaker feature vector. In the second stage, these 12 feature extraction coefficients are used as inputs to feed-forward neural networks. Probabilistic neural network is also proposed for comparison. In contrast to conventional speaker recognition methods that extract features from sentences (or words), the proposed method extracts the features from vowels. Advantages of using vowels include the ability to recognize speakers when only partially-recorded words are available. This may be useful for deaf-mute persons or when the recordings are damaged. Experimental results show that the proposed method succeeds in the speaker verification and identification tasks with high classification rate. This is accomplished with minimum amount of information, using only 12 coefficient features (i.e. vector length) and only one vowel signal, which is the major contribution of this work. The results are further compared to well-known classical algorithms for speaker recognition and are found to be superior.

A Research of Speaker Recognition Based on VQ and MFCC

—Feature extraction is a very important part in speaker recognition system. We proposed and implemented a speaker recognition algorithm based on the VQ and weighted fisher ratio of MFCC. To evaluate performance of this algorithm, we built a small speaker recognition system based on the MATLAB. Compared with the traditional feature selection methods, the characteristic vector obtained via this algorithm has the greatest degree of differentiation in the same dimension. According to the test results, the speaker recognition algorithm we proposed in this paper, can significantly increase the accuracy rate of training and recognition, and reduce the data required by calculation, in the case of keeping a higher recognition rate.

Implementation and optimization of speaker recognition algorithm based on SOPC

Making use of the flexible programmability of SOPC(System On a Programmable Chip) and strong parallel processing ability of FPGA(Field Programmable Gate Array),the speaker recognition algorithm was implemented on FPGA,and the system was optimized in terms of identification speed and accuracy.The principle of speaker recognition algorithm got researched,and according to the characteristics,the SOPC was constructed.It used ping-pong operation to implement voice collection and processing,and used the hardware of FPGA to deal with some time-consuming modules in algorithm so as to quicken the recognition.It also used Genetic Algorithm(GA) to generate template codebook to improve the identification accuracy.Finally,the system realized the function of identity recognition with high real-time quality and accuracy.

Speaker Identification using Spectrograms of Varying Frame Sizes

paper, a text dependent speaker recognition algorithm based on spectrogram is proposed. The spectrograms have been generated using Discrete Fourier Transform for varying frame sizes with 25% and 50% overlap between speech frames. Feature vector extraction has been done by using the row mean vector of the spectrograms. For feature matching, two distance measures, namely Euclidean distance and Manhattan distance have been used. The results have been computed using two databases: a locally created database and CSLU speaker recognition database. The maximum accuracy is 92.52% for an overlap of 50% between speech frames with Manhattan distance as similarity measure.

Analysis of Large-Scale SVM Training Algorithms for Language and Speaker Recognition

This paper compares a set of large scale support vector machine (SVM) training algorithms for language and speaker recognition tasks. We analyze five approaches for training phonetic and acoustic SVM models for language recognition. We compare the performance of these approaches as a function of the training time required by each of them to reach convergence, and we discuss their scalability towards large corpora. Two of these algorithms can be used in speaker recognition to train a SVM that classifies pairs of utterances as either belonging to the same speaker or to two different speakers. Our results show that the accuracy of these algorithms is asymptotically equivalent, but they have different behavior with respect to the time required to converge. Some of these algorithms not only scale linearly with the training set size, but are also able to give their best results after just a few iterations. State-of-the-art performance has been obtained in the female subset of the NIST 2010 Speaker Recognition Evaluation extended core test using a single SVM system.

LPC 켑스트럼 계수와 신경회로망을 사용한 화자인식

본 논문에서는 퍼셉트론 신경회로망과 선형예측부호화 켑스트럼 계수를 사용한 화자인식 알고리즘을 제안한다. 제안하는 화자인식 알고리즘은 입력받은 음성신호에 대해서 유성음 구간을 추출한다. 추출된 유성음 구간에 대하여 선형예측 분석에 의하여 화자의 특성을 가지고 있는 선형예측부호화 켑스트럼 계수를 구한다. 구해진 선형예측부호화 켑스트럼 계수를 분류하기 위하여 이 켑스트럼 계수를 퍼셉트론 신경회로망의 입력으로 사용하여 네트워크의 학습을 수행한다. 본 실험에서는 선형예측부호화 켑스트럼 계수와 신경회로망을 사용하여 본 화자인식 알고리즘이 유효하다는 것을 인식률을 통하여 확인한다. This paper proposes a speaker recognition algorithm using a perceptron neural network and LPC (Linear Predictive Coding) cepstrum coefficients. The proposed algorithm first detects the voiced sections at each frame. Then, the LPC cepstrum coefficients which have speaker characteristics are obtained by the linear predictive analysis for the detected voiced sections. To classify the obtained LPC cepstrum coefficients, a neural network is trained using the LPC cepstrum coefficients. In this experiment, the performance of the proposed algorithm was evaluated using the speech recognition rates based on the LPC cepstrum coefficients and the neural network.

THE VISUAL-AUDIO INTEGRATED RECOGNITION METHOD FOR USER AUTHENTICATION SYSTEM OF PARTNER ROBOTS

Some of noncontact biometric ways have been used for user authentication system of partner robots, such as visual-based recognition methods and speech recognition. However, the methods of visual-based recognition are sensitive to the light noise and speech recognition systems are perturbed to the acoustic environment and sound noise. Inspiring from the human's capability of compensating visual information (looking) with audio information (hearing), a visual-audio integrating method is proposed to deal with the disturbance of light noise and to improve the recognition accuracy. Combining with the PCA-based and 2DPCA-based face recognition, a two-stage speaker recognition algorithm is used to extract useful personal identity information from speech signals. With the statistic properties of visual background noise, the visual-audio integrating method is performed to draw the final decision. The proposed method is evaluated on a public visual-audio dataset VidTIMIT and a partner robot authentication system. The results verified the visual-audio integrating method can obtain satisfied recognition results with strong robustness.