Interface of an Automatic Recognition System for Dysarthric Speech

In this research, a novel feature set is used to automatically segment speech signal. Automatic segmentation is very useful especially for large database. A hybrid features model is created from wavelet packet analysis and mel-scale is used to train Hidden Markov Model (HMM) for phone boundary detection. HMM is implemented using the Hidden Markov Model Toolkit (HTK).The database (Ked-TIMIT) is used for result verifications and Mel Frequency Cepstral Coefficients (MFCC) is used as reference for evaluating the results of the proposed Hybrid model. The results are categorized for vowels, consonants and short phones. Phone duration and start location are used as metrics to evaluate the system success rate. Success rate of 74% is achieved for consonant detection, 72% for vowel detection and 58% for short phone detection. Using the simple metric that relies only on boundary locations but ignoring duration, the achieved results are 92.5% for consonant detection, 90% for vowel detection and 77.5% for short phoneme detection. In addition to boundary detection the proposed hybrid model is utilized to compare newly developed features called Mel scale Best Tree Encoding (Mel-BTE ) to the mostly used popular features MFCC along with all experiments using the same database. The relative results forMel-BTE with respect to MFCC are 94.77% for consonant detection, 87.5% for vowel detection and 93.33% for short phoneme detection.

This paper presents a Bangla (widely used as Bengali) automatic speech recognition system (ASR) by suppressing gender effects. Gender characteristic plays an important role on the performance of ASR. If there is a suppression process that represses the decrease of differences in acoustic-likelihood among categories resulted from gender factors, a robust ASR system can be realized. In the proposed method, we have designed a new ASR incorporating the Local Features (LFs) instead of standard mel frequency cepstral coefficients (MFCCs) as an acoustic feature for Bangla by suppressing the gender effects, which embeds three HMM-based classifiers for corresponding male, female and geneder-independent (GI) characteristics. In the experiments on Bangla speech database prepared by us, the proposed system has achieved a significant improvement of word correct rates (WCRs), word accuracies (WAs) and sentence correct rates (SCRs) in comparison with the method that incorporates Standard MFCCs.

Speech production knowledge has been used to enhance the phonetic representation and the performance of automatic speech recognition (ASR) systems successfully. Representations of speech production make simple explanations for many phenomena observed in speech. These phenomena can not be easily analyzed from either acoustic signal or phonetic transcription alone. One of the most important aspects of speech production knowledge is the use of articulatory knowledge, which describes the smooth and continuous movements in the vocal tract. In this paper, we present a new articulatory model to provide available information for rescoring the speech recognition lattice hypothesis. The articulatory model consists of a feature front-end, which computes a voicing feature based on a spectral harmonics correlation (SHC) function, and a back-end based on the combination of deep neural networks (DNNs) and hidden Markov models (HMMs). The voicing features are incorporated with standard Mel frequency cepstral coefficients (MFCCs) using heteroscedastic linear discriminant analysis (HLDA) to compensate the speech recognition accuracy rates. Moreover, the advantages of two different models are taken into account by the algorithm, which retains deep learning properties of DNNs, while modeling the articulatory context powerfully through HMMs. Mandarin speech recognition experiments show the proposed method achieves significant improvements in speech recognition performance over the system using MFCCs alone.

In this paper, we propose to use syllables as the acoustic units representing speech signals in an automatic speech recognition (ASR) system in order to improve the performance of the automatic recognition of dysarthric speech. The motivation behind using syllables comes from studies of human perception which demonstrate the central role of the syllable played in human perception and generation of speech. To test our proposed approach, a syllable-based speaker-independent HMM-based ASR system was designed using Hidden Markov Model Toolkit (HTK). A series of experiments on dysarthric speech has been carried out using a subset of NEMOURS database. The obtained results show that the relative improvement in the recognition rate using syllables were found to be 8.18% and 15.48% compared to the recognition rates obtained using monophones and triphones, respectively.

Mel Frequency Cepstral Coefficients (MFCCs) and Perceptual linear prediction coefficients (PLPCs) are widely casted nonlinear vocal parameters in majority of the speaker identification, speaker and speech recognition techniques as well in the field of emotion recognition. Post 1980s, significant exertions are put forth on for the progress of these features. Considerations like the usage of appropriate frequency estimation approaches, proposal of appropriate filter banks, and selection of preferred features perform a vital part for the strength of models employing these features. This article projects an overview of MFCC and PLPC features for different speech applications. The insights such as performance metrics of accuracy, background environment, type of data, and size of features are inspected and concise with the corresponding key references. Adding more to this, the advantages and shortcomings of these features have been discussed. This background work will hopefully contribute to floating a heading step in the direction of the enhancement of MFCC and PLPC with respect to novelty, raised levels of accuracy, and lesser complexity.

An additional feature processing algorithm using Non-negative Matrix Factorization (NMF) is proposed to be included during the conventional extraction of Mel-frequency cepstral coefficients (MFCC) for achieving noise robustness in HMM based speech recognition. The proposed approach reconstructs log-Mel filterbank outputs of speech data from a set of building blocks that form the bases of a speech subspace. The bases are learned using the standard NMF of training data. A variation of learning the bases is proposed, which uses histogram equalized activation coefficients during training, to achieve noise robustness. The proposed methods give up to 5.96% absolute improvement in recognition accuracy on Aurora-2 task over a baseline with standard MFCCs, and up to 13.69% improvement when combined with other feature normalization techniques like Histogram Equalization (HEQ) and Heteroscedastic Linear Discriminant Analysis (HLDA).

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.

Frequency masking filtering is introduced in a standard mel frequency cepstral coefficients (MFCC) feature extraction algorithm. It mimics a human masking mechanism to get more robust features when the input speech is distorted by various noises. The AURORA 2.0 database together with HTK speech recognition toolkits are used to evaluate the impact of the frequency masking filtering algorithm at various thresholds. It is shown that with the proper frequency masking coefficients, it can have about 6.59%, 6.01% and 1.20% relative performance improvements over standard MFCC for test A and test B and test C respectively, in clean-condition training. It works well on all eight different noise conditions. It has also proved to be effective when it is combined with other popular noise robust techniques, such as cepstral mean normalization. The proposed frequency masking filtering algorithm is fairly simple and it only requires a very small extra computation load.

Speech activity detection (SAD) on channel transmissions is a critical preprocessing task for speech, speaker and language recognition or for further human analysis. This paper presents a feature combination approach to improve SAD on highly channel degraded speech as part of the Defense Advanced Research Projects Agency’s (DARPA) Robust Automatic Transcription of Speech (RATS) program. The key contribution is the feature combination exploration of different novel SAD features based on pitch and spectro-temporal processing and the standard Mel Frequency Cepstral Coefficients (MFCC) acoustic feature. The SAD features are: (1) a GABOR feature representation, followed by a multilayer perceptron (MLP); (2) a feature that combines multiple voicing features and spectral flux measures (Combo); (3) a feature based on subband autocorrelation (SAcC) and MLP postprocessing and (4) a multiband comb-filter F0 (MBCombF0) voicing measure. We present single, pairwise and all feature combinations, show high error reductions from pairwise feature level combination over the MFCC baseline and show that the best performance is achieved by the combination of all features. Index Terms: speech detection, channel-degraded speech, robust voicing features

This paper reports some of the observations carried out on SUSE database for emotion classification. A comparative study is made to evaluate the performance of Linear Prediction Cepstral Coefficients (LPCCs) and Mel Frequency Cepstral Coefficients (MFCCs) for designing the emotion classification system for word level utterances. The significance of the orders of the coefficients has been carried out during this study. The results obtained using 12th order LPCC and 13th order MFCC are compared with respect to their reduced dimensions of lower orders. A new classification system based on the feature extraction technique using the 2nd, 3rd and 4th order coefficients of both MFCCs and LPCCs is also proposed. This paper compares the accuracy level of both MFCC and LPCC, enabling us to decide which orders of the parameters (both MFCC and LPCC) are more efficient in conveying the emotion for word level utterances. The initial experiments performed at word level utterances reveal that LPCC is more efficient in detecting emotion as compared to MFCC. Further, we noticed that the emotions conveyed in word level utterances are detected more accurately than that in sentence level utterances. The result suggests that word level approach provides better performance for emotion classification as compared to sentence level approach if the system is designed using vocal tract information only.

Feature extraction techniques for input representation to diagnose infant diseases have received significant attention recently. Mel Frequency Cepstral Coefficients (MFCC) is one of the most popular feature extraction techniques due to its representation method being very similar to the human auditory system. The MFCC method for feature extraction depends on several important parameter settings, namely the number of filter banks, and the number of coefficients used in the final representation. These settings affects the way the features are represented, and in turn, affects the performance of the classifier for diagnosis of the disease. In this paper, the Particle Swarm Optimization (PSO) algorithm was used to optimise the parameters of the MFCC feature extraction method for classifying infants with asphyxia. The extracted MFCC features were then used to train several MLP classifiers over different initialization values. The accuracy of these classifiers was then used to guide the PSO optimization. Our results show that the optimization of MFCC computation using PSO yielded 93.9% accuracy, an improvement of 1.45% over typical MFCC parameter settings using the same classifier.

The short-term Fourier transform magnitude spectra (STFT-MS) computed from the dysarthric speech deviates nonlinearly from the normal speech in different frequency bands depending on underlying sound units. This discriminating information can be captured by segmenting the STFT-MS into different frequency bands following the power spectra of board categories of sound units. Motivated by this observation in this study, we have computed the cepstral coefficients by analyzing the STFT-MS in 0–500 Hz, 500–2000 Hz, 2000–4000 Hz, and 4000 – 8000Hz, respectively for 16 kHz sampled speech data. Each of the selected frequency bands is analyzed by using a 30 channel Mel filterbank. The log filterbank energies computed for each sub-band are then polled together and discrete cosine transform (DCT) is applied to compute the cepstral coefficients, here termed as sub-band enhanced Mel frequency cepstral coefficients (SE-MFCC). The i-vector based dysarthric intelligibility assessment system reported in this study shows that the SEMFCC outperforms the conventional Mel frequency cepstral coefficients (MFCC), and the cepstral coefficients computed using inverse-Mel filterbank (IMFCC), and linear filterbank (LFCC). The score level combination of SE-MFCC with the MFCC further improves the overall performance.

Background/Objectives: Speech produced by individuals with hearing loss differs notably from that of normal-hearing (NH) individuals. Although cochlear implants (CIs) provide sufficient auditory input to support speech acquisition and control, there remains considerable variability in speech intelligibility among CI users. As a result, speech produced by CI talkers often exhibits distinct acoustic characteristics compared to that of NH individuals. Methods: Speech data were obtained from eight cochlear-implant (CI) and eight normal-hearing (NH) talkers, while electroencephalogram (EEG) responses were recorded from 11 NH listeners exposed to the same speech stimuli. Support Vector Machine (SVM) classifiers employing 3-fold cross-validation were evaluated using classification accuracy as the performance metric. This study evaluated the efficacy of Support Vector Machine (SVM) algorithms using four kernel functions (Linear, Polynomial, Gaussian, and Radial Basis Function) to classify speech produced by NH and CI talkers. Six acoustic features—Log Energy, Zero-Crossing Rate (ZCR), Pitch, Linear Predictive Coefficients (LPC), Mel-Frequency Cepstral Coefficients (MFCCs), and Perceptual Linear Predictive Cepstral Coefficients (PLP-CC)—were extracted. These same features were also extracted from electroencephalogram (EEG) recordings of NH listeners who were exposed to the speech stimuli. The EEG analysis leveraged the assumption of quasi-stationarity over short time windows. Results: Classification of speech signals using SVMs yielded the highest accuracies of 100% and 94% for the Energy and MFCC features, respectively, using Gaussian and RBF kernels. EEG responses to speech achieved classification accuracies exceeding 70% for ZCR and Pitch features using the same kernels. Other features such as LPC and PLP-CC yielded moderate to low classification performance. Conclusions: The results indicate that both speech-derived and EEG-derived features can effectively differentiate between CI and NH talkers. Among the tested kernels, Gaussian and RBF provided superior performance, particularly when using Energy and MFCC features. These findings support the application of SVMs for multimodal classification in hearing research, with potential applications in improving CI speech processing and auditory rehabilitation.

Speaker identification refers to the process of recognizing human voice using artificial intelligence techniques. Speaker identification technologies are widely applied in voice authentication, security and surveillance, electronic voice eavesdropping, and identity verification. In the speaker identification process, extracting discriminative and salient features from speaker utterances is an important task to accurately identify speakers. Various features for speaker identification have been recently proposed by researchers. Most studies on speaker identification have utilized short-time features, such as perceptual linear predictive (PLP) coefficients and Mel frequency cepstral coefficients (MFCC), due to their capability to capture the repetitive nature and efficiency of signals. Various studies have shown the effectiveness of MFCC features in correctly identifying speakers. However, the performances of these features degrade on complex speech datasets, and therefore, these features fail to accurately identify speaker characteristics. To address this problem, this study proposes a novel fusion of MFCC and time-based features (MFCCT), which combines the effectiveness of MFCC and time-domain features to improve the accuracy of text-independent speaker identification (SI) systems. The extracted MFCCT features were fed as input to a deep neural network (DNN) to construct the speaker identification model. Results showed that the proposed MFCCT features coupled with DNN outperformed existing baseline MFCC and time-domain features on the LibriSpeech dataset. In addition, DNN obtained better classification results compared with five machine learning algorithms that were recently utilized in speaker recognition. Moreover, this study evaluated the effectiveness of one-level and two-level classification methods for speaker identification. The experimental results showed that two-level classification presented better results than one-level classification. The proposed features and classification model for identifying a speaker can be widely applied to different types of speaker datasets.

Mobile technology is very fast growing and incredible, yet there are not much technology development and improvement for Deaf‐mute peoples. Existing mobile applications use sign language as the only option for communication with them. Before our article, no such application (app) that uses the disrupted speech of Deaf‐mutes for the purpose of social connectivity exists in the mobile market. The proposed application, named as vocalizer to mute (V2M), uses automatic speech recognition (ASR) methodology to recognize the speech of Deaf‐mute and convert it into a recognizable form of speech for a normal person. In this work mel frequency cepstral coefficients (MFCC) based features are extracted for each training and testing sample of Deaf‐mute speech. The hidden Markov model toolkit (HTK) is used for the process of speech recognition. The application is also integrated with a 3D avatar for providing visualization support. The avatar is responsible for performing the sign language on behalf of a person with no awareness of Deaf‐mute culture. The prototype application was piloted in social welfare institute for Deaf‐mute children. Participants were 15 children aged between 7 and 13 years. The experimental results show the accuracy of the proposed application as 97.9%. The quantitative and qualitative analysis of results also revealed that face‐to‐face socialization of Deaf‐mute is improved by the intervention of mobile technology. The participants also suggested that the proposed mobile application can act as a voice for them and they can socialize with friends and family by using this app.