Power Normalized Cepstral Coefficients Research Articles

The Power-Normalized Cepstral Coefficient (PNCC) for convolutional neural networks-based robust speech command recognition

While implementations of speech recognition grow rapidly in recent years and are slowly being integrated into our daily devices, the problem of noise robustness is still a challenging task, even with the recent advancement of deep learning technologies for speech recognition. The presence of noise may cause a mismatch between training, which is performed in clean conditions, and noisy testing conditions. This paper proposes a method to extract features for speech recognition by employing features derived under the power law scale, i.e., the Power-Normalized Cepstral Coefficient (PNCC). The power-law can provide better compression in low-energy regions so that it is not sensitive when the speech signal is distorted by noise. The features are implemented on speech recognition based on Convolutional Neural Networks (CNNs). The experiments were carried out by TensorFlow’s Speech Command Dataset mixed with various signal-to-noise ratio to evaluate the method. The experimental findings indicate that the accuracy ranges from 81% to 86%.

Speech Recognition Algorithm in a Noisy Environment Based on Power Normalized Cepstral Coefficient and Modified Weighted-KNN

Speech Recognition Algorithm in a Noisy Environment Based on Power Normalized Cepstral Coefficient and Modified Weighted-KNN

A high-performance speech BioHashing retrieval algorithm based on audio segmentation

As one of the research hotspots in the field of speech recognition, content-based speech retrieval algorithms can detect speech information with the same content features, which improves computer intelligence while reducing labor costs, and thus have been widely used. Although most of the current speech content retrieval algorithms can guarantee excellent retrieval performance for small-scale speech retrieval work, the performance of the above algorithms is greatly reduced under the constraints of large speech data storage space and high content redundancy. In order to solve the above problems, a high-performance speech BioHashing retrieval algorithm based on audio segmentation is proposed in this paper. The algorithm is divided into an offline pre-processing phase and an online retrieval phase, The offline pre-processing stage converts the speech data into BioHashing sequences with speech content characteristics. In this process, first of all, the Power-Normalized Cepstral Coefficients (PNCC) features of the speech data are extracted and biometric templates with single mapping keys are constructed according to the PNCC features, obtaining BioHashing sequences. Then, slice the original speeches into short-time audio segments according to the proposed audio segmentation algorithm, and the hash reconstruction operation is performed on the BioHashing sequences to obtain the reconstructed Hashing sequences for online retrieval. The online search phase responds to the users' query requests, just find the hash index that matches the query hash sequence from the BioHashing index table, and will the standardized editing distance (SED) to the closest 1 value corresponding to the hash index as the retrieval result back to the user. The experimental results show that the reconstructed hash sequences obtained after removing the silent redundant segments have better robustness and discrimination. Moreover, the algorithm achieves 100% retrieval accuracy for the original speech clips, and the average retrieval time is only 0.0157 s, which shows that the algorithm has good retrieval performance and can meet the needs of speech retrieval in various environments.

Design and Implementation of Machine Tool Life Inspection System Based on Sound Sensing.

The main causes of damage to industrial machinery are aging, corrosion, and the wear of parts, which affect the accuracy of machinery and product precision. Identifying problems early and predicting the life cycle of a machine for early maintenance can avoid costly plant failures. Compared with other sensing and monitoring instruments, sound sensors are inexpensive, portable, and have less computational data. This paper proposed a machine tool life cycle model with noise reduction. The life cycle model uses Mel-Frequency Cepstral Coefficients (MFCC) to extract audio features. A Deep Neural Network (DNN) is used to understand the relationship between audio features and life cycle, and then determine the audio signal corresponding to the aging degree. The noise reduction model simulates the actual environment by adding noise and extracts features by Power Normalized Cepstral Coefficients (PNCC), and designs Mask as the DNN's learning target to eliminate the effect of noise. The effect of the denoising model is improved by 6.8% under Short-Time Objective Intelligibility (STOI). There is a 3.9% improvement under Perceptual Evaluation of Speech Quality (PESQ). The life cycle model accuracy before denoising is 76%. After adding the noise reduction system, the accuracy of the life cycle model is increased to 80%.

Developing sequentially trained robust Punjabi speech recognition system under matched and mismatched conditions

Development of a native language robust ASR framework is very challenging as well as an active area of research. Although an urge for investigation of effective front-end as well as back-end approaches are required for tackling environment differences, large training complexity and inter-speaker variability in achieving success of a recognition system. In this paper, four front-end approaches: mel-frequency cepstral coefficients (MFCC), Gammatone frequency cepstral coefficients (GFCC), relative spectral-perceptual linear prediction (RASTA-PLP) and power-normalized cepstral coefficients (PNCC) have been investigated to generate unique and robust feature vectors at different SNR values. Furthermore, to handle the large training data complexity, parameter optimization has been performed with sequence-discriminative training techniques: maximum mutual information (MMI), minimum phone error (MPE), boosted-MMI (bMMI), and state-level minimum Bayes risk (sMBR). It has been demonstrated by selection of an optimal value of parameters using lattice generation, and adjustments of learning rates. In proposed framework, four different systems have been tested by analyzing various feature extraction approaches (with or without speaker normalization through Vocal Tract Length Normalization (VTLN) approach in test set) and classification strategy on with or without artificial extension of train dataset. To compare each system performance, true matched (adult train and test—S1, child train and test—S2) and mismatched (adult train and child test—S3, adult + child train and child test—S4) systems on large adult and very small Punjabi clean speech corpus have been demonstrated. Consequently, gender-based in-domain data augmented is used to moderate acoustic and phonetic variations throughout adult and children’s speech under mismatched conditions. The experiment result shows that an effective framework developed on PNCC + VTLN front-end approach using TDNN-sMBR-based model through parameter optimization technique yields a relative improvement (RI) of 40.18%, 47.51%, and 49.87% in matched, mismatched and gender-based in-domain augmented system under typical clean and noisy conditions, respectively.

Isolated word recognition based on PNCC with different classifiers in a noisy environment

Speech Recognition is one of the important methods to control robotics, mobile applications, automobile, etc. Nevertheless, still, the Speech recognition rate needs to be improved especially in noisy environments. The main goal of this paper is to investigate the best features for isolated word recognition between the most common features extraction technique Mel Frequency Cepstral Coefficient MFCC and noise immunity features extraction technique Power Normalized Cepstral Coefficient PNCC, with the aid of four classifiers, Weighted KNN, Medium KNN, Linear SVM, Quadratic SVM in different levels of noise. The database that is used for this study is Speech Commands Data Set v0.01 which is containing 64,727 audio files to get a good estimation of the recognition rate. The algorithm based on MFCC is recommended for Speaker dependent speech recognition a little bit better (1.1%) than the algorithm based on PNCC in a noise-free environment. Nevertheless, the proposed algorithm based on PNCC is more efficient than the algorithm based on MFCC for Speaker independent speech recognition, and overall Recognition accuracy. Moreover, it has more immunity to noise and it is more accurate than MFCC by 8.7% on average with already trained speakers and 18.528% on average with new trained speakers’ tests. Finally, the proposed algorithm based on the PNCC feature extraction with aid of Weighted KNN is more accurate than other algorithms for Speaker independent speech recognition and overall Recognition accuracy. Moreover, it has more immunity to noise.

Combined i-Vector and Extreme Learning Machine Approach for Robust Speaker Identification and Evaluation with SITW 2016, NIST 2008, TIMIT Databases

In this article, a novel combined i-vector and an Extreme Learning Machine (ELM) is proposed for speaker identification. The ELM is chosen because it is fast to train and has a universal approximator property. Four combinations of features based on Mel Frequency Cepstral Coefficient and Power Normalized Cepstral Coefficient are used. Besides, seven fusion methods are exploited. The system is evaluated with three different databases, namely: the SITW 2006, NIST 2008, and the TIMIT database. This work employs the 2016 SITW database for the first time for speaker identification using the integration between the ELM and i-vector approach. From each database, 120 speakers with 1200 speech utterances are used (overall 360 speakers with 3600 speech utterances). Furthermore, comprehensive evaluations are exploited with a wide range of realistic background noise types (Stationary noise AWGN and Non-Stationary Noise types) with the handset effect. The proposed system is compared with the Gaussian Mixture Model-Universal Background Model (GMM-UBM) and other states of the art approaches. The results show that the i-vector method outperforms the GMM-UBM approach and other state- of-the-art methods under specific conditions, and that fusion techniques can be used to improve robustness to noise and handset effects.

Improved Power Normalized Cepstrum Coefficient Based on Wavelet Packet Decomposition for Trunk Borer Detection in Harsh Acoustic Environment

The sound-detection method of trunk borer is a very promising method in the field of forestry prevention and control of trunk borers. However, the detection accuracy of commonly used algorithms often decreases sharply in the case of noise reverberation interference. In practical applications, the sound monitoring of trunk borers often takes place in a harsh acoustic environment. To solve this problem, we intend to introduce methods which are effective in other related acoustic fields. Unfortunately, most of the methods are not suitable for acoustic detection of trunk borers and perform extremely poorly. After trying various methods, we found that Power-Normalized Cepstral Coefficients (PNCC) performed well in some cases, while it did not in others. This is due to the difference between speech and trunk borer sound. Therefore, an improved anti-noise PNCC based on wavelet package is proposed. The dmey wavlet system always obtains the best performance. We collected the audio of the following five dry borer pests for testing. They are red palm weevil, mountain pine beetle, red necked longicorn, Asian longhorn beetle and citrus longhorn beetle. In the experimental part, we used genetic algorithm-support vector machine (GA-SVM) as a classifier to compare Mel Cepstral Coefficients (MFCC), which are the most common methods in the field of audio detection of trunk borer, PNCC and improved PNCC in a variety of noise environments. The results showed that, compared with other methods, the newly proposed method can often achieve better results. The above experiments take the audio clips made of clear pest sound mixed noise. In order to further verify the effectiveness of the method, we designed another experiment with a harsh outdoor acoustic environment. We found that the proposed method achieved 88% accuracy and the traditional PNCC achieved 78% accuracy. However, the Mel cepstrum coefficient completely lost its ability to distinguish. In sum, the proposed PNCC based on wavelet packet decomposition can be used as a detection method for trunk borer in the harsh acoustic environment. This method has many advantages, including simple extraction and strong robustness to noise. Combined with cheap audio acquisition equipment, this method can effectively improve the early warning ability of forestry borer pests.

Speech Intelligibility Enhancement Algorithm Based on Multi-Resolution Power-Normalized Cepstral Coefficients (MRPNCC) for Digital Hearing Aids

Speech intelligibility enhancement in noisy environments is still one of the major challenges for hearing impaired in everyday life. Recently, Machine-learning based approaches to speech enhancement have shown great promise for improving speech intelligibility. Two key issues of these approaches are acoustic features extracted from noisy signals and classifiers used for supervised learning. In this paper, features are focused. Multi-resolution power-normalized cepstral coefficients (MRPNCC) are proposed as a new feature to enhance the speech intelligibility for hearing impaired. The new feature is constructed by combining four cepstrum at different time–frequency (T–F) resolutions in order to capture both the local and contextual information. MRPNCC vectors and binary masking labels calculated by signals passed through gammatone filterbank are used to train support vector machine (SVM) classifier, which aim to identify the binary masking values of the T–F units in the enhancement stage. The enhanced speech is synthesized by using the estimated masking values and wiener filtered T–F unit. Objective experimental results demonstrate that the proposed feature is superior to other comparing features in terms of HIT-FA, STOI, HASPI and PESQ, and that the proposed algorithm not only improves speech intelligibility but also improves speech quality slightly. Subjective tests validate the effectiveness of the proposed algorithm for hearing impaired.

Pitch and noise normalized acoustic feature for children's ASR

In this work, we have analyzed the pitch robustness of the recently reported power normalized cepstral coefficient (PNCC) feature for noise robust children's speech recognition. The PNCC feature is intended to suppress various types of common additive noise present in speech data. The PNCC feature is noted to be susceptible to pitch variations. In order to normalize the pitch effect on PNCC, a pitch base spectral normalization step is incorporated in the PNCC feature extraction process. By the inclusion of pitch normalization stage, the pitch-robustness of PNCC is enhanced without loss of noise suppression capability. The efficacy of the proposed pitch normalized PNCC (PN-PNCC) feature is investigated using acoustic modeling based on recurrent structures, such as gated recurrent units (GRU), and light gated recurrent units (LiGRU). The PN-PNCC feature is noted to yield performance improvement over PNCC for adults' and children's speech recognition under clean as well as the noisy test condition.

Active Sonar Target Classification with Power-Normalized Cepstral Coefficients and Convolutional Neural Network

Detection and classification of unidentified underwater targets maneuvering in complex underwater environments are critical for active sonar systems. In previous studies, many detection methods were applied to separate targets from the clutter using signals that exceed a preset threshold determined by the sonar console operator. This is because the high signal-to-noise ratio target has enough feature vector components to separate. However, in a real environment, the signal-to-noise ratio of the received target does not always exceed the threshold. Therefore, a target detection algorithm for various target signal-to-noise ratio environments is required; strong clutter energy can lead to false detection, while weak target signals reduce the probability of detection. It also uses long pulse repetition intervals for long-range detection and high ambient noise, requiring classification processing for each ping without accumulating pings. In this study, a target classification algorithm is proposed that can be applied to signals in real underwater environments above the noise level without a threshold set by the sonar console operator, and the classification performance of the algorithm is verified. The active sonar for long-range target detection has low-resolution data; thus, feature vector extraction algorithms are required. Feature vectors are extracted from the experimental data using Power-Normalized Cepstral Coefficients for target classification. Feature vectors are also extracted with Mel-Frequency Cepstral Coefficients and compared with the proposed algorithm. A convolutional neural network was employed as the classifier. In addition, the proposed algorithm is to be compared with the result of target classification using a spectrogram and convolutional neural network. Experimental data were obtained using a hull-mounted active sonar system operating on a Korean naval ship in the East Sea of South Korea and a real maneuvering underwater target. From the experimental data with 29 pings, we extracted 361 target and 3351 clutter data. It is difficult to collect real underwater target data from the real sea environment. Therefore, the number of target data was increased using the data augmentation technique. Eighty percent of the data was used for training and the rest was used for testing. Accuracy value curves and classification rate tables are presented for performance analysis and discussion. Results showed that the proposed algorithm has a higher classification rate than Mel-Frequency Cepstral Coefficients without affecting the target classification by the signal level. Additionally, the obtained results showed that target classification is possible within one ping data without any ping accumulation.

Adaptive wavelet thresholding with robust hybrid features for text-independent speaker identification system

The robustness of speaker identification system over additive noise channel is crucial for real-world applications. In speaker identification (SID) systems, the extracted features from each speech frame are an essential factor for building a reliable identification system. For clean environments, the identification system works well; in noisy environments, there is an additive noise, which is affect the system. To eliminate the problem of additive noise and to achieve a high accuracy in speaker identification system a proposed algorithm for feature extraction based on speech enhancement and a combined features is presents. In this paper, a wavelet thresholding pre-processing stage, and feature warping (FW) techniques are used with two combined features named power normalized cepstral coefficients (PNCC) and gammatone frequency cepstral coefficients (GFCC) to improve the identification system robustness against different types of additive noises. Universal Background Model Gaussian Mixture Model (UBM-GMM) is used for features matching between the claim and actual speakers. The results showed performance improvement for the proposed feature extraction algorithm of identification system comparing with conventional features over most types of noises and different SNR ratios.

Unsupervised feature selection and NMF de-noising for robust Speech Emotion Recognition

Speech feature fusion is the most commonly used phenomenon for improving the accuracy in Speech Emotion Recognition (SER). But in this, there is a disadvantage of increasing the complexity in SER system in terms of processing time. Besides this, some of the features could be redundant and do not contribute for SER and lead to incorrect emotion prediction and reduction in SER accuracy. To overcome this problem, in this paper, unsupervised feature selection is applied to the feature set with the combination of INTERSPEECH 2010 paralinguistic features, Gammatone Cepstral Coefficients (GTCC) and Power Normalized Cepstral Coefficients (PNCC). The Feature Selection with Adaptive Structure Learning (FSASL), Unsupervised Feature Selection with Ordinal Locality (UFSOL) and a novel Subset Feature Selection (SuFS) algorithm are the feature dimension reduction techniques used to acquire better SER performance in this work. The proposed SER system is analyzed in both clean and noisy environments. The EMO-DB and IEMOCAP emotion databases are considered for evaluating the proposed SER performance. For noise analysis, the clean speech is corrupted with different noises of Aurora noise database and white Gaussian noise at different Signal to Noise Ratio (SNR) levels from −5dB to 20 dB. Support Vector Machine (SVM) classifier with linear and Radial Basis Function (RBF) kernels using 10-fold cross-validation and hold-out validation is used in this analysis with classification accuracy and computation time as the performance metrics. The results show that the proposed SER system outperforms the baseline SER system as well as many of the existing literature works both in clean and noisy conditions. For SNR levels >15 dB, the proposed SER system in presence of different noises performs same as the SER in clean environments. Whereas, for lower SNRs <15 dB the performance is likely to be reduced. Therefore, to overcome this drawback and improve the SER performance in noisy conditions, a dense Non-Negative Matrix Factorization (denseNMF) method is adopted for de-noising the noisy speech signal prior to SER achieving noise robustness.

ROBUST HYBRID FEATURES BASED TEXT INDEPENDENT SPEAKER IDENTIFICATION SYSTEM OVER NOISY ADDITIVE CHANNEL

Robustness of speaker identification systems over additive noise is crucial for real-world applications. In this paper, two robust features named Power Normalized Cepstral Coefficients (PNCC) and Gammatone Frequency Cepstral Coefficients (GFCC) are combined together to improve the robustness of speaker identification system over different types of noise. Universal Background Model Gaussian Mixture Model (UBM-GMM) is used as a feature matching and a classifier to identify the claim speakers. Evaluation results show that the proposed hybrid feature improves the performance of identification system when compared to conventional features over most types of noise and different signal-to-noise ratios.

Comparison of feature extraction and normalization methods for speaker recognition using grid-audiovisual database

&lt;p&gt;&lt;span lang="EN-GB"&gt;In this paper, different feature extraction and feature normalization methods are investigated for speaker recognition. With a view to give a good representation of acoustic speech signals, Power Normalized Cepstral Coefficients (PNCCs) and Mel Frequency Cepstral Coefficients (MFCCs) are employed for feature extraction. Then, to mitigate the effect of linear channel, Cepstral Mean-Variance Normalization (CMVN) and feature warping are utilized. The current paper investigates Text-independent speaker identification system by using 16 coefficients from both the MFCCs and PNCCs features. Eight different speakers are selected from the GRID-Audiovisual database with two females and six males. The speakers are modeled using the coupling between the Universal Background Model and Gaussian Mixture Models (GMM-UBM) in order to get a fast scoring technique and better performance. The system shows 100% in terms of speaker identification accuracy. The results illustrated that PNCCs features have better performance compared to the MFCCs features to identify females compared to male speakers. Furthermore, feature wrapping reported better performance compared to the CMVN method. &lt;/span&gt;&lt;/p&gt;

Efficient Feature Extraction Algorithms to Develop an Arabic Speech Recognition System

This paper studies three feature extraction methods, Mel-Frequency Cepstral Coefficients (MFCC), Power-Normalized Cepstral Coefficients (PNCC), and Modified Group Delay Function (ModGDF) for the development of an Automated Speech Recognition System (ASR) in Arabic. The Support Vector Machine (SVM) algorithm processed the obtained features. These feature extraction algorithms extract speech or voice characteristics and process the group delay functionality calculated straight from the voice signal. These algorithms were deployed to extract audio forms from Arabic speakers. PNCC provided the best recognition results in Arabic speech in comparison with the other methods. Simulation results showed that PNCC and ModGDF were more accurate than MFCC in Arabic speech recognition.

Enhanced Automatic Speech Recognition System Based on Enhancing Power-Normalized Cepstral Coefficients

Many new consumer applications are based on the use of automatic speech recognition (ASR) systems, such as voice command interfaces, speech-to-text applications, and data entry processes. Although ASR systems have remarkably improved in recent decades, the speech recognition system performance still significantly degrades in the presence of noisy environments. Developing a robust ASR system that can work in real-world noise and other acoustic distorting conditions is an attractive research topic. Many advanced algorithms have been developed in the literature to deal with this problem; most of these algorithms are based on modeling the behavior of the human auditory system with perceived noisy speech. In this research, the power-normalized cepstral coefficient (PNCC) system is modified to increase robustness against the different types of environmental noises, where a new technique based on gammatone channel filtering combined with channel bias minimization is used to suppress the noise effects. The TIDIGITS database is utilized to evaluate the performance of the proposed system in comparison to the state-of-the-art techniques in the presence of additive white Gaussian noise (AWGN) and seven different types of environmental noises. In this research, one word is recognized from a set containing 11 possibilities only. The experimental results showed that the proposed method provides significant improvements in the recognition accuracy at low signal to noise ratios (SNR). In the case of subway noise at SNR = 5 dB, the proposed method outperforms the mel-frequency cepstral coefficient (MFCC) and relative spectral (RASTA)–perceptual linear predictive (PLP) methods by 55% and 47%, respectively. Moreover, the recognition rate of the proposed method is higher than the gammatone frequency cepstral coefficient (GFCC) and PNCC methods in the case of car noise. It is enhanced by 40% in comparison to the GFCC method at SNR 0dB, while it is improved by 20% in comparison to the PNCC method at SNR −5dB.

Ia-PNCC: Noise Processing Method for Underwater Target Recognition Convolutional Neural Network

Underwater target recognition is a key technology for underwater acoustic countermeasure. And how to classify and recognize underwater targets according to the noise information of underwater targets has been a hot topic in the field of underwater acoustic signals. In this paper, the deep learning model is applied in the underwater target recognition and the improved anti-noise Power-Normalized Cepstral Coefficients (ia-PNCC) is proposed based on PNCC oriented to underwater target noise features. In this coefficient, multitaper and normalized Gammatone filter group are used to improve the anti-noise capacity of PNCC in underwater target recognition, and it is combined with the convolutional neural network to recognize the underwater target. The experiment results show that the acoustic feature presented by ia-PNCC is of higher anti-noise capacity and more adaptive to the underwater target recognition model of convolutional neural network. Compared with the combination of convolutional neural network with single acoustic feature such as MFCC (Mel-scale frequency cepstral coefficients) or LPCC (Linear Prediction PLP) and so on, the combination of ia-PNCC with convolutional neural network improves the underwater target recognition accuracy greatly.

An investigation on the degradation of different features extracted from the compressed American English speech using narrowband and wideband codecs

Speech coding facilitates speech compression without perceptual loss that results in the elimination or deterioration of both speech and speaker specific features used for a wide range of applications like automatic speaker and speech recognition, biometric authentication, prosody evaluations etc. The present work investigates the effect of speech coding in the quality of features which include Mel Frequency Cepstral Coefficients, Gammatone Frequency Cepstral Coefficients, Power-Normalized Cepstral Coefficients, Perceptual Linear Prediction Cepstral Coefficients, Rasta-Perceptual Linear Prediction Cepstral Coefficients, Residue Cepstrum Coefficients and Linear Predictive Coding-derived cepstral coefficients extracted from codec compressed speech. The codecs selected for this study are G.711, G.729, G.722.2, Enhanced Voice Services, Mixed Excitation Linear Prediction and also three codecs based on compressive sensing frame work. The analysis also includes the variation in the quality of extracted features with various bit-rates supported by Enhanced Voice Services, G.722.2 and compressive sensing codecs. The quality analysis of extracted epochs, fundamental frequency and formants estimated from codec compressed speech was also performed here. In the case of various features extracted from the output of selected codecs, the variation introduced by Mixed Excitation Linear Prediction codec is the least due to its unique method for the representation of excitation. In the case of compressive sensing based codecs, there is a drastic improvement in the quality of extracted features with the augmentation of bit rate due to the waveform type coding used in compressive sensing based codecs. For the most popular Code Excited Linear Prediction codec based on Analysis-by-Synthesis coding paradigm, the impact of Linear Predictive Coding order in feature extraction is investigated. There is an improvement in the quality of extracted features with the order of linear prediction and the optimum performance is obtained for Linear Predictive Coding order between 20 and 30, and this varies with gender and statistical characteristics of speech. Even though the basic motive of a codec is to compress single voice source, the performance of codecs in multi speaker environment is also studied, which is the most common environment in majority of the speech processing applications. Here, the multi speaker environment with two speakers is considered and there is an augmentation in the quality of individual speeches with increase in diversity of mixtures that are passed through codecs. The perceptual quality of individual speeches extracted from the codec compressed speech is almost same for both Mixed Excitation Linear Prediction and Enhanced Voice Services codecs but regarding the preservation of features, the Mixed Excitation Linear Prediction codec has shown a superior performance over Enhanced Voice Services codec.

THE AUTOMATIC SPEAKER RECOGNITION SYSTEM OF CRITICAL USE CLASSIFIER OPTIMIZATION

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 as a result of the factor analysis, their visual representation for the use of the convolution neural network classifier, and optimizing its architecture for the needs of ASRSCU. Method. Measures are proposed to optimize the speaker recognition procedure of the ASRSCU, for which the optimal way of informative features representation and the method of increasing their informativeness are theoretically justified, the topology and measures for increasing of 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 that will be analyzed by a convolution neural network, an optimal method of factor analysis (a sparse main components analyzing method) to reduce of the features vector length while preserving its informativeness, an improved topology of the convolution neural network in which the Gabor filters are integrated in to the convolution layer, which allows them to optimize their parameters during the neural network training process, and in a fully connected layer a deep neural network with a bottleneck layer is used, whose weights after training are uses as inputs for the GMM/HMM control classifier. Results. Methods of representation and optimization of the speaker’s individual features, methods for their visual presentation and improvement 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 convolution neural network to the noisy input phonograms proved to be higher than the results of an ordinary convolution neural network and a deep neural network. 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 that provide the most variable individual features from the speech signal for speaker recognition are determined empirically.