Improve Speech Recognition Performance Research Articles

Multi-view Self-supervised Learning and Multi-scale Feature Fusion for Automatic Speech Recognition

To address the challenges of the poor representation capability and low data utilization rate of end-to-end speech recognition models in deep learning, this study proposes an end-to-end speech recognition model based on multi-scale feature fusion and multi-view self-supervised learning (MM-ASR). It adopts a multi-task learning paradigm for training. The proposed method emphasizes the importance of inter-layer information within shared encoders, aiming to enhance the model’s characterization capability via the multi-scale feature fusion module. Moreover, we apply multi-view self-supervised learning to effectively exploit data information. Our approach is rigorously evaluated on the Aishell-1 dataset and further validated its effectiveness on the English corpus WSJ. The experimental results demonstrate a noteworthy 4.6%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document} reduction in character error rate, indicating significantly improved speech recognition performance . These findings showcase the effectiveness and potential of our proposed MM-ASR model for end-to-end speech recognition tasks.

Multichannel AV-wav2vec2: A Framework for Learning Multichannel Multi-Modal Speech Representation

Self-supervised speech pre-training methods have developed rapidly in recent years, which show to be very effective for many near-field single-channel speech tasks. However, far-field multichannel speech processing is suffering from the scarcity of labeled multichannel data and complex ambient noises. The efficacy of self-supervised learning for far-field multichannel and multi-modal speech processing has not been well explored. Considering that visual information helps to improve speech recognition performance in noisy scenes, in this work we propose the multichannel multi-modal speech self-supervised learning framework AV-wav2vec2, which utilizes video and multichannel audio data as inputs. First, we propose a multi-path structure to process multi-channel audio streams and a visual stream in parallel, with intra-, and inter-channel contrastive as training targets to fully exploit the rich information in multi-channel speech data. Second, based on contrastive learning, we use additional single-channel audio data, which is trained jointly to improve the performance of multichannel multi-modal representation. Finally, we use a Chinese multichannel multi-modal dataset in real scenarios to validate the effectiveness of the proposed method on audio-visual speech recognition (AVSR), automatic speech recognition (ASR), visual speech recognition (VSR) and audio-visual speaker diarization (AVSD) tasks.

Retracted: Use Brain-Like Audio Features to Improve Speech Recognition Performance

Retracted: Use Brain-Like Audio Features to Improve Speech Recognition Performance

Robust speech recognition based on deep learning for sports game review

Abstract To verify the feasibility of robust speech recognition based on deep learning in sports game review. In this paper, a robust speech recognition model is built based on the generative adversarial network GAN algorithm according to the deep learning model. And the loss function, optimization function and noise reduction front-end are introduced in the model to achieve the optimization of speech extraction features through denoising process to ensure that accurate speech review data can be derived even in the game scene under noisy environment. Finally, the experiments are conducted to verify the four directions of the model algorithm by comparing the speech features MFCC, FBANK and WAVE. The experimental results show that the speech recognition model trained by the GSDNet model algorithm can reach 89% accuracy, 56.24% reduction of auxiliary speech recognition word error rate, 92.61% accuracy of speech feature extraction, about 62.19% reduction of training sample data volume, and 94.75% improvement of speech recognition performance in the speech recognition task under noisy environment. It shows that the robust speech recognition based on deep learning can be applied to sports game reviews, and also can provide accurate voice review information from the noisy sports game scene, and also broaden the application area for deep learning models.

Improving Speech Recognition Performance in Noisy Environments by Enhancing Lip Reading Accuracy.

The current accuracy of speech recognition can reach over 97% on different datasets, but in noisy environments, it is greatly reduced. Improving speech recognition performance in noisy environments is a challenging task. Due to the fact that visual information is not affected by noise, researchers often use lip information to help to improve speech recognition performance. This is where the performance of lip recognition and the effect of cross-modal fusion are particularly important. In this paper, we try to improve the accuracy of speech recognition in noisy environments by improving the lip reading performance and the cross-modal fusion effect. First, due to the same lip possibly containing multiple meanings, we constructed a one-to-many mapping relationship model between lips and speech allowing for the lip reading model to consider which articulations are represented from the input lip movements. Audio representations are also preserved by modeling the inter-relationships between paired audiovisual representations. At the inference stage, the preserved audio representations could be extracted from memory by the learned inter-relationships using only video input. Second, a joint cross-fusion model using the attention mechanism could effectively exploit complementary intermodal relationships, and the model calculates cross-attention weights on the basis of the correlations between joint feature representations and individual modalities. Lastly, our proposed model achieved a 4.0% reduction in WER in a -15 dB SNR environment compared to the baseline method, and a 10.1% reduction in WER compared to speech recognition. The experimental results show that our method could achieve a significant improvement over speech recognition models in different noise environments.

Differential Effects of Binaural Pitch Fusion Range on the Benefits of Voice Gender Differences in a "Cocktail Party" Environment for Bimodal and Bilateral Cochlear Implant Users.

Some cochlear implant (CI) users are fitted with a CI in each ear ("bilateral"), while others have a CI in one ear and a hearing aid in the other ("bimodal"). Presently, evaluation of the benefits of bilateral or bimodal CI fitting does not take into account the integration of frequency information across the ears. This study tests the hypothesis that CI listeners, especially bimodal CI users, with a more precise integration of frequency information across ears ("sharp binaural pitch fusion") will derive greater benefit from voice gender differences in a multi-talker listening environment. Twelve bimodal CI users and twelve bilateral CI users participated. First, binaural pitch fusion ranges were measured using the simultaneous, dichotic presentation of reference and comparison stimuli (electric pulse trains for CI ears and acoustic tones for HA ears) in opposite ears, with reference stimuli fixed and comparison stimuli varied in frequency/electrode to find the range perceived as a single sound. Direct electrical stimulation was used in implanted ears through the research interface, which allowed selective stimulation of one electrode at a time, and acoustic stimulation was used in the non-implanted ears through the headphone. Second, speech-on-speech masking performance was measured to estimate masking release by voice gender difference between target and maskers (VGRM). The VGRM was calculated as the difference in speech recognition thresholds of target sounds in the presence of same-gender or different-gender maskers. Voice gender differences between target and masker talkers improved speech recognition performance for the bimodal CI group, but not the bilateral CI group. The bimodal CI users who benefited the most from voice gender differences were those who had the narrowest range of acoustic frequencies that fused into a single sound with stimulation from a single electrode from the CI in the opposite ear. There was no similar voice gender difference benefit of narrow binaural fusion range for the bilateral CI users. The findings suggest that broad binaural fusion reduces the acoustical information available for differentiating individual talkers in bimodal CI users, but not for bilateral CI users. In addition, for bimodal CI users with narrow binaural fusion who benefit from voice gender differences, bilateral implantation could lead to a loss of that benefit and impair their ability to selectively attend to one talker in the presence of multiple competing talkers. The results suggest that binaural pitch fusion, along with an assessment of residual hearing and other factors, could be important for assessing bimodal and bilateral CI users.

Improving N-Best Rescoring in Under-Resourced Code-Switched Speech Recognition Using Pretraining and Data Augmentation

In this study, we present improvements in N-best rescoring of code-switched speech achieved by n-gram augmentation as well as optimised pretraining of long short-term memory (LSTM) language models with larger corpora of out-of-domain monolingual text. Our investigation specifically considers the impact of the way in which multiple monolingual datasets are interleaved prior to being presented as input to a language model. In addition, we consider the application of large pretrained transformer-based architectures, and present the first investigation employing these models in English-Bantu code-switched speech recognition. Our experimental evaluation is performed on an under-resourced corpus of code-switched speech comprising four bilingual code-switched sub-corpora, each containing a Bantu language (isiZulu, isiXhosa, Sesotho, or Setswana) and English. We find in our experiments that, by combining n-gram augmentation with the optimised pretraining strategy, speech recognition errors are reduced for each individual bilingual pair by 3.51% absolute on average over the four corpora. Importantly, we find that even speech recognition at language boundaries improves by 1.14% even though the additional data is monolingual. Utilising the augmented n-grams for lattice generation, we then contrast these improvements with those achieved after fine-tuning pretrained transformer-based models such as distilled GPT-2 and M-BERT. We find that, even though these language models have not been trained on any of our target languages, they can improve speech recognition performance even in zero-shot settings. After fine-tuning on in-domain data, these large architectures offer further improvements, achieving a 4.45% absolute decrease in overall speech recognition errors and a 3.52% improvement over language boundaries. Finally, a combination of the optimised LSTM and fine-tuned BERT models achieves a further gain of 0.47% absolute on average for three of the four language pairs compared to M-BERT. We conclude that the careful optimisation of the pretraining strategy used for neural network language models can offer worthwhile improvements in speech recognition accuracy even at language switches, and that much larger state-of-the-art architectures such as GPT-2 and M-BERT promise even further gains.

A Preliminary Study of Robust Speech Feature Extraction Based on Maximizing the Probability of States in Deep Acoustic Models

This study proposes a novel robust speech feature extraction technique to improve speech recognition performance in noisy environments. This novel method exploits the information provided by the original acoustic model in the automatic speech recognition (ASR) system to learn a deep neural network that converts the original speech features. This deep neural network is trained to maximize the posterior accuracy of the state sequences of acoustic models with respect to the speech feature sequences. Compared with the robustness methods that retrain or adapt acoustic models, the new method has the advantages of less computation load and faster training. In the experiments conducted on the medium-vocabulary TIMIT database and task, the presented method provides lower word error rates than the unprocessed baseline and speech-enhancement-based techniques. These results indicate that the presented method is promising and worth further developing.

Parameter tuning of time-frequency masking algorithms for reverberant artifact removal within the cochlear implant stimulus

Cochlear implant recipients struggle to understand speech in reverberant environments. To restore speech perception, artifacts due to reverberant reflections can be removed from the cochlear implant stimulus by applying a matrix of gain values, a technique referred to as time-frequency masking. In this study, two common time-frequency masking strategies are implemented within cochlear implant processing, either introducing complete retention or deletion of stimulus components using a binary mask or continuous attenuation of stimulus components using a ratio mask. Parameters of each masking strategy control the level of attenuation imposed by the gain values. In this study, we perceptually tune the parameters of the masking strategy to determine a balance between speech retention and artifact removal. We measure the intelligibility of reverberant signals mitigated by each strategy with speech recognition testing in normal-hearing listeners using vocoding as a simulation of cochlear implant perception. For both masking strategies, we find parameterizations that maximize the intelligibility of the mitigated signals. At the best-performing parameterizations, binary-masked reverberant signals yield larger intelligibility improvements than ratio-masked signals. The results provide a perceptually optimized objective for the removal of reverberant artifacts from cochlear implant stimuli, facilitating improved speech recognition performance for cochlear implant recipients in reverberant environments.

DLD: An Optimized Chinese Speech Recognition Model Based on Deep Learning

Speech recognition technology has played an indispensable role in realizing human-computer intelligent interaction. However, most of the current Chinese speech recognition systems are provided online or offline models with low accuracy and poor performance. To improve the performance of offline Chinese speech recognition, we propose a hybrid acoustic model of deep convolutional neural network, long short-term memory, and deep neural network (DCNN-LSTM-DNN, DLD). This model utilizes DCNN to reduce frequency variation and adds a batch normalization (BN) layer after its convolutional layer to ensure the stability of data distribution, and then use LSTM to effectively solve the gradient vanishing problem. Finally, the fully connected structure of DNN is utilized to efficiently map the input features into a separable space, which is helpful for data classification. Therefore, leveraging the strengths of DCNN, LSTM, and DNN by combining them into a unified architecture can effectively improve speech recognition performance. Our model was tested on the open Chinese speech database THCHS-30 released by the Center for Speech and Language Technology (CSLT) of Tsinghua University, and it was concluded that the DLD model with 3 layers of LSTM and 3 layers of DNN had the best performance, reaching 13.49% of words error rate (WER).

Effects of Wireless Remote Microphone on Speech Recognition in Noise for Hearing Aid Users in China.

Objective: This study was aimed at evaluating improvements in speech-in-noise recognition ability as measured by signal-to-noise ratio (SNR) with the use of wireless remote microphone technology. These microphones transmit digital signals via radio frequency directly to hearing aids and may be a valuable assistive listening device for the hearing-impaired population of Mandarin speakers in China.Methods: Twenty-three adults (aged 19–80 years old) and fourteen children (aged 8–17 years old) with bilateral sensorineural hearing loss were recruited. The Mandarin Hearing in Noise Test was used to test speech recognition ability in adult subjects, and the Mandarin Hearing in Noise Test for Children was used for children. The subjects’ perceived SNR was measured using sentence recognition ability at three different listening distances of 1.5, 3, and 6 m. At each distance, SNR was obtained under three device settings: hearing aid microphone alone, wireless remote microphone alone, and hearing aid microphone and wireless remote microphone simultaneously.Results: At each test distance, for both adult and pediatric groups, speech-in-noise recognition thresholds were significantly lower with the use of the wireless remote microphone in comparison with the hearing aid microphones alone (P < 0.05), indicating better SNR performance with the wireless remote microphone. Moreover, when the wireless remote microphone was used, test distance had no effect on speech-in-noise recognition for either adults or children.Conclusion: Wireless remote microphone technology can significantly improve speech recognition performance in challenging listening environments for Mandarin speaking hearing aid users in China.

Linguistic-Coupled Age-to-Age Voice Translation to Improve Speech Recognition Performance in Real Environments

We address a low-performance problem of the elderly in automatic speech recognition (ASR) through feature adaptation agnostic to the ASR. Most of the datasets for speech recognition models consist of datasets collected from adult speakers. Consequently, the majority of commercial speech recognition systems typically tend to perform well on adult speakers. In other words, the limited diversity of speakers in the training datasets yields unreliable performance for minority (e.g., elderly) speakers due to the infeasible acquisition of training data. In response, this paper suggests a neural network-based voice conversion framework to enhance speech recognition of the minority. To this end, we propose a voice translation model including an unsupervised phonology clustering to extract linguistic information to fit the minority’s speech to a current acoustic model frame. Our proposal is a spectral feature adaptation method that can be placed in front of any commercial or open ASR system, avoiding directly modifying the speech recognizer. The experimental results and analysis demonstrate the effectiveness of our proposed method through improvement in elderly speech recognition accuracy.

Confusion analysis in phoneme based speech recognition in Hindi

Phoneme recognition is an essential step in the development of a speech recognition system (SRS), as phonemes are fundamental building blocks in a spoken language. This research work aimed to present phoneme recognition with systematic confusion analysis for the Hindi language. The accuracy of phoneme recognition is the foundation for developing an efficient SRS. Therefore, the systematic confusion analysis for phoneme recognition is essential to improve speech recognition performance. Experiments conducted on Continuous Hindi speech corpus for phoneme recognition with speaker-dependent mode using Hidden Markov Model (HMM) based tool kit HTK. Feature extraction technique Perceptual Linear Predictive Coefficient (PLP) was used with five states Monophones HMM model. Tests were performed for exploring the recognition of Hindi vowels and consonants. Confusion matrices were presented for both vowels and consonants with analysis and possible solutions. During systematic analysis, the vowels were divided into front, middle, and back vowels while consonants were categorized based on place of articulation and manner of articulation. Research findings show that some Hindi phonemes have significant effects on speech recognition. The investigations also reveal that some Hindi phonemes are mostly confused, and some phonemes have more deletions and insertions. The research further demonstrates that the words made of less number of phonemes show more insertion errors. It was also found that most of the Hindi sentences end with some specific words. These particular words can be used to reduce the search place in language modeling for improving speech recognition. The research findings can be utilized to enhance the performance of the speech recognition system by selecting suitable feature extraction techniques and classification techniques for phonemes. The outcome of the research can also be used to develop improved pronunciation dictionaries and designing the text for developing phonetically balanced speech corpus for improvement in speech recognition. Experimental results show an average corrected recognition score of 70% for vowel class and consonant categories, the maximum average corrected recognition score of 94% was obtained with palatal sounds, and the lowest average corrected recognition score of 54% was achieved with liquid sounds. The comparative analysis of the presented work was made to similar existing works.

General properties of auditory spectro-temporal receptive fields.

The analyses of more than 2000 spectro-temporal receptive fields (STRFs) obtained from the ferret primary auditory cortex revealed their dominant encoding properties along time and frequency domains. Results showed that the peak responses of the STRFs were roughly aligned along the time axis, and enhanced low modulation components of the signal around 3 Hz. On the contrary, the peaks of the STRF population along the frequency axis varied and were distributed widely. Further analyses revealed that there were some general properties along the frequency axis around the best frequency of each STRF. The STRFs enhanced modulation frequencies around 0.25 cycles/octave. These findings are consistent with techniques that have been reported in the literature to result in improvements in speech recognition performance, highlighting the significant potential of biological insights into the design of better machines.

Knowledge Transferability Between the Speech Data of Persons With Dysarthria Speaking Different Languages for Dysarthric Speech Recognition

In this paper, we present an end-to-end speech recognition system for Japanese persons with articulation disorders resulting from athetoid cerebral palsy. Because their utterance is often unstable or unclear, speech recognition systems struggle to recognize their speech. Recent deep learning-based approaches have exhibited promising performance. However, these approaches require a large amount of training data, and it is difficult to collect sufficient data from such dysarthric people. This paper proposes a transfer learning method that transfers two types of knowledge corresponding to the different datasets: the language-dependent (phonetic and linguistic) characteristic of unimpaired speech and the language-independent characteristic of dysarthric speech. The former is obtained from Japanese non-dysarthric speech data, and the latter is obtained from non-Japanese dysarthric speech data. In the proposed method, we pre-train a model using Japanese non-dysarthric speech and non-Japanese dysarthric speech, and thereafter, we fine-tune the model using the target Japanese dysarthric speech. To handle the speech data of the two different languages in one model, we employ language-specific decoder modules. Experimental results indicate that our proposed approach can significantly improve speech recognition performance compared with other approaches that do not use additional speech data.

형태소 발음변이를 고려한 음성인식 단위의 성능*

This paper proposes a method to improve speech recognition performance by extracting various pronunciations of the pseudo-morpheme unit from an eojeol unit corpus and generating a new recognition unit considering pronunciation variations. In the proposed method, we first align the pronunciation of the eojeol units and the pseudo-morpheme units, and then expand the pronunciation dictionary by extracting the new pronunciations of the pseudo-morpheme units at the pronunciation of the eojeol units. Then, we propose a new recognition unit that relies on pronunciation by tagging the obtained phoneme symbols according to the pseudo-morpheme units. The proposed units and their extended pronunciations are incorporated into the lexicon and language model of the speech recognizer. Experiments for performance evaluation are performed using the Korean speech recognizer with a trigram language model obtained by a 100 million pseudo-morpheme corpus and an acoustic model trained by a multi-genre broadcast speech data of 445 hours. The proposed method is shown to reduce the word error rate relatively by 13.8% in the news-genre evaluation data and by 4.5% in the total evaluation data.

Speech Recognition in Noise in Adults and Children Who Speak English or Chinese as Their First Language.

Speech recognition of individuals who are listening to a nonnative language is significantly degraded in the presence of background noise and may be influenced by proficiency, age of acquisition, language experience, and daily use of the nonnative language. The purpose of this study is to examine and compare speech recognition in noise performance across test conditions with varying signal-to-noise ratios (SNRs) as well as the presence of vocal and spatial cues in listeners who speak American English as a native language or Mandarin Chinese as a native language. Self-rated English proficiency and experience were collected for native Mandarin Chinese speakers to determine its relationship to performance on the test measures. A cross-sectional repeated measures design was used for the study. Four groups of participants were included in the study. The adult groups consisted of 25 adults who speak native English and 25 adults who speak native Mandarin Chinese with English as an additional language. The pediatric groups consisted of 16 children who speak native English and 16 children who speak native Mandarin Chinese with English as an additional language. Percent correct speech recognition in noise was assessed at three SNRs (-3, 0, +3 dB) using the adult or pediatric versions of the AzBio sentence test. The Listening in Spatialized Noise-Sentence (LiSN-S) test was used to determine the effect of providing spatial and vocal cues on the speech recognition in noise performance of the groups of participants. The data for each age group and test measure were analyzed with a repeated measures analysis of variance. Correlation analyses were performed to examine relationships between English proficiency and experience on performance across the speech recognition test conditions. Analysis of the data from the adult or pediatric AzBio sentence test identified a significant effect of native language for adults but no significant effect for children. The higher SNRs yielded better performance for all listeners. On the LiSN-S test, results for the adult and pediatric groups were similar and showed significantly better performance for the native English speakers in every test condition. The demographic and language characteristics that most affected speech recognition performance across the test measures included the length of time the person lived in the United States, the age of English acquisition, the number of minutes per day English was spoken by the participant, and the self-rated English proficiency. The findings in this study highlight the importance and benefit of higher SNRs as well as the provision of vocal and spatial cues for improving speech recognition performance in noise of adult and pediatric listeners who speak Mandarin Chinese as a native language.

BoneBridge implantation in patients with single-sided deafness resulting from vestibular schwannoma resection: objective and subjective benefit evaluations

Background: The BoneBridge could rehabilitate hearing for patients with single-sided deafness (SSD).Objectives: To evaluate the objective and subjective benefits of BoneBridge implantation in patients after vestibular Schwannoma resection and to explore the factors affecting the benefits.Material and methods: We prospectively enrolled all 15 patients implanted with BoneBridge after VS resection from January to June 2017. The primary outcome was the ability to hear in noisy conditions. The secondary outcomes were the soft-band BoneBridge try-on rate, the frequency of BB use, the sound source localization test result, and questionnaire measures of quality of life (QoL).Results: Patients showed better speech recognition ability in the presence of noise with the BoneBridge. The BoneBridge provided no help in sound localization, although most patients reported subjective sound localization benefits. The results of QoL questionnaires showed significant satisfaction with BoneBridge implantation. The unilateral hearing deprivation duration and high education levels had significant impacts on the subjective benefits of patients.Conclusions: The BoneBridge could improve speech recognition performance in complex auditory backgrounds, as well as QoL, especially in patients with short unilateral hearing deprivation durations and high education levels.Significance: The BoneBridge is an effective hearing aid for single-sided deafness patients after VS removal.

Shifts in Maximum Audiovisual Integration with Age.

Listeners attempting to understand speech in noisy environments rely on visual and auditory processes, typically referred to as audiovisual processing. Noise corrupts the auditory speech signal and listeners naturally leverage visual cues from the talker's face in an attempt to interpret the degraded auditory signal. Studies of speech intelligibility in noise show that the maximum improvement in speech recognition performance (i.e., maximum visual enhancement or VEmax), derived from seeing an interlocutor's face, is invariant with age. Several studies have reported that VEmax is typically associated with a signal-to-noise (SNR) of -12dB; however, few studies have systematically investigated whether the SNR associated with VEmax changes with age. We investigated if VEmax changes as a function of age, whether the SNR at VEmax changes as a function of age, and what perceptual/cognitive abilities account for or mediate such relationships. We measured VEmax on a nongeriatric adult sample (N=64) ranging in age from 20 to 59 years old. We found that VEmax was age-invariant, replicating earlier studies. No perceptual/cognitive measures predicted VEmax, most likely due to limited variance in VEmax scores. Importantly, we found that the SNR at VEmax shifts toward higher (quieter) SNR levels with increasing age; however, this relationship is partially mediated by working memory capacity, where those with larger working memory capacities (WMCs) can identify speech under lower (louder) SNR levels than their age equivalents with smaller WMCs. The current study is the first to report that individual differences in WMC partially mediate the age-related shift in SNR at VEmax.

Rank-1 constrained Multichannel Wiener Filter for speech recognition in noisy environments

Multichannel linear filters, such as the Multichannel Wiener Filter (MWF) and the Generalized Eigenvalue (GEV) beamformer are popular signal processing techniques which can improve speech recognition performance. In this paper, we present an experimental study on these linear filters in a specific speech recognition task, namely the CHiME-4 challenge, which features real recordings in multiple noisy environments. Specifically, the rank-1 MWF is employed for noise reduction and a new constant residual noise power constraint is derived which enhances the recognition performance. To fulfill the underlying rank-1 assumption, the speech covariance matrix is reconstructed based on eigenvectors or generalized eigenvectors. Then the rank-1 constrained MWF is evaluated with alternative multichannel linear filters under the same framework, which involves a Bidirectional Long Short-Term Memory (BLSTM) network for mask estimation. The proposed filter outperforms alternative ones, leading to a 40% relative Word Error Rate (WER) reduction compared with the baseline Weighted Delay and Sum (WDAS) beamformer on the real test set, and a 15% relative WER reduction compared with the GEV-BAN method. The results also suggest that the speech recognition accuracy correlates more with the Mel-frequency cepstral coefficients (MFCC) feature variance than with the noise reduction or the speech distortion level.