Performance Analysis of Isolated Speech Recognition Technique Using MFCC and Cross-Correlation

The aim of the study was to inspect the acoustic properties and sleep characteristics of a preapneic snoring sound. The feasibility of forecasting upcoming respiratory events by snoring sound was also investigated. Participants with habitual snoring or a heavy breathing sound during sleep were recruited consecutively. Polysomnography was conducted, and snoring-related breathing sound was recorded simultaneously. Acoustic features and sleep features were extracted from 30-second samples, and a machine learning algorithm was used to establish 2 prediction models. A total of 74 eligible participants were included. Model 1, tested by 5-fold cross-validation, achieved an accuracy of 0.92 and an area under the curve of 0.94 for respiratory event prediction. Model 2, with acoustic features and sleep information tested by Leave-One-Out cross-validation, had an accuracy of 0.78 and an area under the curve of 0.80. Sleep position was found to be the most important among all sleep features contributing to the performance of the 2 models. Preapneic sound presented unique acoustic characteristics, and snoring-related breathing sound could be deployed as a real-time apneic event predictor. The models, combined with sleep information, serve as a promising tool for an early warning system to forecast apneic events. Wang B, Yi X, Gao J, etal. Real-time prediction of upcoming respiratory events via machine learning using snoring sound signal. J Clin Sleep Med. 2021;17(9):1777-1784.

This paper examines the performance improvement via combining feature vectors for speaker verification. As feature vectors, we use LPC (linear predictive coding) spectral envelope (LPCSE), Mel-frequency cepstral coefficients (MFCC) and piecewise LPC (pLPC) pole distribution (pLPCPD). Here, pLPCPD is the feature which we have introduced and developed, and this paper shows a relationship of the underlying pLPC speech signal model for speaker verification with ARMA and AR models for speech production. We execute experiments of speaker verification with the combinations of the features. We examine the experimental results and show that (1) the smallest AER (average error rate) is achieved by MFCC followed in order by pLPCPD and LPCSE, (2) the AER achieved by LPCSE, MFCC and LPCSE+MFCC (combination of LPCSE and MFCC) is reduced by combining pLPCPD, (3) there are conflicting features in LPCSE, MFCC and pLPCPD to increase the AER, and the smallest AER is obtained by MFCC+pLPCPD among the combinations of LPCSE, MFCC and pLPCPD. We also investigate the relationship between pLPCPD and the aperiodicity used for vocal cord excitation in recent vocoders, which seems to explain the above result (2).

Due to the differences in speech speed and intonation among different users, traditional speech recognition methods are difficult to accurately evaluate the user's speech, resulting in training failure. Therefore, this paper is dedicated to researching and developing a Chinese speech evaluation and intelligent training system based on Dynamic Time Warping (DTW). First, the overall design of the system is carried out to clarify the functional requirements of the system. Secondly, the speech signal is preprocessed, and the input speech signal is filtered, sampled, quantized, and other operations are performed. Then the speech signal is framed and windowed, and it is divided into multiple short-term stable speech frames. The Mel-frequency cepstral coefficient (MFCC) is used to extract feature parameters that can reflect the speech characteristics from the preprocessed speech frames. Then, the DTW algorithm is used to perform pattern matching on the extracted feature parameters, and the cumulative distance between the test speech sequence and the template speech sequence is calculated. The degree of similarity between the test speech and the template speech is judged by comparing the size of the cumulative distance. Subsequently, the evaluation module evaluates the speech quality of the test speech according to the cumulative distance output by the DTW algorithm processing module, and gives the corresponding evaluation results. Finally, the intelligent training module provides users with personalized training guidance based on the evaluation results and the training needs of users. The system's evaluation results were consistent with manual evaluations at an average of 97.4%, and users' MOS and STOI scores improved after training with the system. The results of this study can improve the voice quality of users and promote the popularization and application of voice interaction technology.

Prior studies in the automatic classification of voice quality have mainly studied the use of the acoustic speech signal as input. Recently, a few studies have been carried out by jointly using both speech and neck surface accelerometer (NSA) signals as inputs, and by extracting mel-frequency cepstral coefficients (MFCCs) and glottal source features. This study examines simultaneously-recorded speech and NSA signals in the classification of voice quality (breathy, modal, and pressed) using features derived from three self-supervised pre-trained models (wav2vec2-BASE, wav2vec2-LARGE, and HuBERT) and using a support vector machine (SVM) as well as convolutional neural networks (CNNs) as classifiers. Furthermore, the effectiveness of the pre-trained models is compared in feature extraction between glottal source waveforms and raw signal waveforms for both speech and NSA inputs. Using two signal processing methods (quasi-closed phase (QCP) glottal inverse filtering and zero frequency filtering (ZFF)), glottal source waveforms are estimated from both speech and NSA signals. The study has three main goals: (1) to study whether features derived from pre-trained models improve classification accuracy compared to conventional features (spectrogram, mel-spectrogram, MFCCs, i-vector, and x-vector), (2) to investigate which of the two modalities (speech vs. NSA) is more effective as input in the classification task with pre-trained model-based features, and (3) to evaluate whether the deep learning-based CNN classifier can enhance the classification accuracy in comparison to the SVM classifier. The results revealed that the use of the NSA input showed better classification performance compared to the speech signal. Between the features, the pre-trained model-based features showed better classification accuracies, both for speech and NSA inputs compared to the conventional features. The two classifiers performed equally well for all the pre-trained model-based features for both speech and NSA signals. It was also found that the HuBERT features performed better than the wav2vec2-BASE and wav2vec2-LARGE features for both speech and NSA inputs. In particular, when compared to the conventional features, the HuBERT features showed an absolute accuracy improvement of 3%–6% for speech and NSA signals in the classification of voice quality.

—The research on speech recognition systems currently focuses on the analysis of robust speech recognition systems. When the speech signals are combined with noise, the recognition system becomes distracted, struggling to identify the speech sounds. Therefore, the development of a robust speech recognition system continues to be carried out. The principle of a robust speech recognition system is to eliminate noise from the speech signals and restore the original information signals. In this paper, researchers conducted a frequency domain analysis on one stage of the Mel Frequency Cepstral Coefficients (MFCC) process, the Fast Fourier Transform (FFT), in children's speech recognition system. The FTT analysis in the feature extraction process determined the effect of frequency value characteristics utilized in the FFT output on the noise disruption. The analysis method was designed into three scenarios based on the value of the employed FFT points. The differences between scenarios were based on the number of shared FFT points. All FFT points were divided into four, three, and two parts in the first, second, and third scenarios, respectively. This study utilized children's speech data from the isolated TIDIGIT English digit corpus. As comparative data, the noise was added manually to simulate real-world conditions. The results showed that using a particular frequency portion following the scenario designed on MFCC affected the recognition system performance, which was relatively significant on the noisy speech data. The designed method in the scenario 3 (C1) version generated the highest accuracy, exceeded the accuracy of the conventional MFCC method. The average accuracy in the scenario 3 (C1) method increased by 1% more than all the tested noise types. Using various noise intensity values (SNR), the testing process indicates that scenario 3 (C1) generates a higher accuracy than conventional MFCC in all tested SNR values. It proves that the selection of specific frequency utilized in MFCC feature extraction significantly affects the recognition accuracy in a noisy speech.

The Mel-Frequency Cepstral Coefficients (MFCC) feature can be cast-off in speaker recognition. The process of feature extraction of the speech signal using Mel-Frequency Cepstral Coefficients (MFCC) feature vectors will generate an acoustic speech signal. Locality Sensitive Hashing (LSH) is frequently used as a classifier for Big Data related problems. In this research, we proposed a new model based on MFCC and LSH to integrate into speaker recognition model. The main returns of our newly proposed model are to get robustness, effective and accurate results in comparison with MFCC+GMM, LPCC+GMM and MFCC+PNN models. This model also contributes to the literature of Big Data. In this model, first, we extract the MFCC features from the wave file then we applied LSH classifier on extracted feature to transform into hash-table. Finally, the hash-tables of train and test wave files are compared and obtained 92.66% speaker recognition accuracy. We compared the accuracy ratio of proposed model with other traditional models namely MFCC+GMM, MFCC+PNN, and LPCC+GMM. Experimental results show that proposed model is more accurate and robust than traditional models and good for speaker recognition.

This paper presents the implementation of a practical voice recognition system using MATLAB (R2014b) to secure a given user’s system so that only the user may access it. Voice recognition systems have two phases, training and testing. During the training phase, the characteristic features of the speaker are extracted from the speech signal and stored in a database. In the testing phase, the stored audio features of the test voice sample are compared with the voice samples in the database and determined if a match exists. For this research, Mel Frequency Cepstral Coefficients (MFCCs) were chosen to represent the feature vectors of the user’s voice as it accurately simulates the behavior of the human ear. This characteristic of the MFCCs makes them an excellent measure of speaker characteristics. The feature matching process is then performed by subjecting the MFCCs to vector quantization using the LBG (Linde-Buzo-Gray) algorithm. In practical scenarios, noise is a major factor that adversely influences a voice recognition system. The paper addresses this issue by utilizing spectral subtraction to remove environmental noise affecting the speech signal thereby increasing the robustness of the system.

Significance of analytic phase of speech signals in speaker verification

The vast number of information packed into the human speech signal makes analysis a tough undertaking. This intricacy is notably noticeable in tasks like speaker recognition, especially when it comes to gender distinction. In this paper, we address this issue by conducting a thorough examination of the effectiveness of various speech features, namely Pitch, Formant Frequency, MFCC (Mel Frequency Cepstral Coefficients), and Chroma, in the context of gender identification in the Tai Language, which is spoken by the Tai people of Assam. In this study, we use machine learning (SVM, KNN, Decision Tree, Neural Network) to analyze speech features (Pitch, Formant Frequency, MFCC, Chroma) for gender identification in the Tai language spoken by the Tai people of Assam. Our results show that MFCC consistently outperforms other features, delivering the highest accuracy rates across all approaches. This demonstrates MFCC’s ability to extract gender information from Tai Language speech signals, suggesting more accurate gender identification systems. Beyond gender identification, our study extends voice analysis in linguistics and improves the application of spoken language data, allowing for improved communication systems and linguistic insights. In summary, our findings highlight the critical significance of MFCC in gender identification in the Tai language, with ramifications that extend far beyond its local context, promising advances in voice analysis and improving our understanding of language and human communication.

This paper presents a convolutional neural network-based pathological voice detection system using speech and electroglottographic (EGG) signals. Speech signals have been popularly used to detect voice pathology. Recently, the EGG signals have drawn considerable attention from researchers in this field. They argued that the EGG signals could detect the vocal fold vibration more accurately than speech signals and hence can be considered more appropriate for voice pathology detection. This work investigates the effectiveness of the EGG and speech signals in detecting pathological voices using sustained vowel (“/a/”) samples collected from the Saarbrücken Voice Database (SVD). The Mel frequency cepstral coefficients (MFCCs) extracted from the speech and EGG samples are employed as discerning features for this investigation. The results show that the proposed system achieves a higher accuracy (more than 23%) in identifying dysphonic voices from healthy ones with speech signals.

Automatic speech recognition (ASR) system is used to recognize the text transcript from the given speech signal. Such speech signal can contain either isolated words or large vocabulary continuous speech (LVCS). Isolated words can be recognized with high accuracy in clean environment, but recognizing continuous words involves various parameters like speech corpus, speaker, environment noise, etc., that affects the accuracy of automatic speech recognition system directly. In the proposed work, hybrid feature extraction technique combines the perceptual linear predictive (PLP) and mel frequency cepstral coefficient (MFCC) to improve the accuracy of ASR in noisy environment. Voice activity and detection (VAD)-based frame dropping is used for improving the phonemes modeling by removing the pauses and distorted elements from the given speech signal. The proposed hybrid model with VAD is implemented by using self-generated speech corpus and shows relatively 12% increase in recognition rate compared with the state-of-the-art methodology.KeywordsAutomatic speech recognition (ASR)Perceptual linear predictive (PLP)Mel frequency cepstral coefficient (MFCC)Voice activity and detection (VAC)

In this paper, we proposed a new set of speech feature parameters based on formant structure information. The speech signal is first divided into a gammatone filterbank, and then the Teager energy signal of each sub-band is extracted according to Teager Energy Operator. Two different energy separation algorithms, DESA-1 and DESA-2, are applied for obtain the instantaneous amplitude and frequency envelope of the formants, respectively. Finally, the feature vector is constructed by the amplitude and frequency information of the formants. The motivation of developing this new feature is that the formant location information is a quite distinct speech representation but seldom applied into speech recognition system before, and the conventional feature parameters such as mel-frequency cepstral coefficients (MFCC) and linear prediction cepstral coefficients (LPCC) do not explicitly model spectral peak information which is very important clue to identify the different phones. A Mandarin digit string recognition task is performed for evaluating the performance of the proposed feature parameter. The recognition results show an improved speech recognition performance compared to the conventional MFCC and LPCC.

Spoken language identification is a field of research that is already being done by many people. There are many techniques proposed for doing speech processing, such as Support Vector Machines, Gaussian Mixture Models, Decision Trees, and others. This paper will use the system using the Mel-Frequency Cepstral Coefficient (MFCC) features of speech input signal, use Random Forest (RF), Gaussian Mixture Model (GMM), and K-Nearest Neighbor (KNN) as a classifier, use the 3s, 10s, and 30s as scoring method, and use dataset that consists of Javanese, Sundanese, and Minang languages which are traditional languages from Indonesia. K-Nearest Neighbor has 98.88% of accuracy for 30s of speech and followed by Random Forest that has 95.55% of accuracy for 30s of speech, GMM has 82.24% of accuracy.

The extraction of representative characteristics from voice signals, is the basic step for any application where speech signal processing is performed. In this paper, a comparative analysis of the Mel frequency cepstral coefficients, extracted from samples of voiced and silent speech is exposed. The classical methodology for extracting Mel frequency cepstral coefficient is evaluated to show the implementation steps and the robustness of such kind of features representing in a unique form the statistical and representative information of time varying signals. To achieve this comprehensive evaluation, first, a contextualization in theoretical terms of the methodology is performed to show the mathematical basis for the coefficient extraction. Each step of the process is analyzed in terms of the representative information of the coefficients. Therefore an evaluation in comparative terms against the previous developments for data interpretation is done. The coefficients are extracted using the classical implementation of the Mel frequency cepstral coefficient feature extraction process in a free software framework implemented in Matlab®. The results show the robustness of the Mel frequency cepstral coefficient algorithm in the characterization of voiced and silent speech signals, which represents an advantage for the construction of any kind of speech processing systems.

In this paper, we apply speech and audio processing techniques to bird vocalizations and for the classification of birds found in the lower Himalayan regions. Mel frequency cepstral coefficients (MFCC) are extracted from each recording. As a result, the recordings are now represented as varying length sets of feature vectors. Dynamic kernel based support vector machines (SVMs) and deep neural networks (DNNs) are popularly used for the classification of such varying length patterns obtained from speech signals. In this work, we propose to use dynamic kernel based SVMs and DNNs for classification of bird calls represented as sets of feature vectors. Results of our studies show that both approaches give comparable performance.