Power-Normalized Cepstral Coefficients (PNCC) for robust speech recognition

The present work investigates the robustness of Power Normalized Cepstral Coefficients (PNCC) for Language identification (LID) from noisy speech. Though the state of the art vocal tract features like mel frequency cepstral coefficients (MFCC) give good recognition accuracy in clean environments, the performance degrades drastically when the signal to noise ratio decreases. In this work, experiments have been carried out on IITKGP-MLILSC speech database. Gaussian mixture model (GMM) is used to building the language models. We have used NOISEX-92 database to add synthetic noise at different SNR levels. We have also compared the recognition accuracy of two systems, one developed using MFCCs and and the other using PNCCs. Finally, we have shown that PNCC features are more robust to noise.

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

Punjabi is a regional language having variant pronunciations and ranging tones. Therefore developing a robust speech recognition system for such language is today's vital need. Conventional research practices lay stress on using basic extraction techniques like Frequency Cepstral Coefficients (MFCC), Perceptual Linear Prediction (PLP) etc. This papers subject the application of a novel Feature Extraction Technique called Power-Normalized Cepstral Coefficients (PNCC) on connected words in Punjabi Automatic Speech Recognition. This particular technique is an extension to conventional MFCC as MFCC uses mel filter bank and traditional log non linearity on other end the proposed technique uses gammatone filter bank and power law non linearity in acoustic modelling phase. As phoneme sounds (tones) are focused, phone based modelling has been enrolled using 34 phones for training 158 words. The phones are used to break each phoneme word into frames based on the sound produced. Training was done in noise-free environment using 16 speakers and noisy environments with 12 speakers. HTK (Hidden Markov Model toolkit) is used to build acoustic model hatching 8 HMM models for training. Each speaker sheered the speech corpus for number of times to increase the redundancy of the system hence, blooming the recognition rate. Performance analysis was done in noisy-free and noisy environments indulging 3 speakers giving 83.05% accuracy in noisy-free and 71.92% accuracy in noisy environment.

Local spectral variability features for speaker verification

Cosine distance similarities with a set of reference speakers, cosine distance features (CDF), with a backend support vector machine classifier (CDF-SVM) have been explored in our earlier studies for improving the performance of speaker verification systems. Subsequently, we also investigated on its effectiveness in improving the noise robustness of speaker verification systems. In this work, we study how the performance of CDF-SVM systems can be further improved by weighting the feature vectors using latent semantic information (LSI) technique. We use mel frequency cepstral coefficients (MFCC), power normalized cepstral coefficients (PNCC), or delta spectral cepstral coefficients (DSCC) for deriving CDF. Experimental results on the female part of short2-short3 trials of NIST speaker recognition evaluation dataset show that the proposed weighted CDF-SVM system outperforms the baseline i-vector with cosine distance scoring (i-CDS), i-vector with a backend SVM classifier (i-SVM) and CDF-SVM systems. Finally, we fused the weighted CDF-SVM with i-CDS and the performance of the combined system was evaluated under different stationary and non-stationary additive noise test conditions. It was seen that the noise robustness of the fused weighted CDF-SVM+i-CDS system is significantly better than the individual systems and the fused CDF-SVM+i-CDS of our earlier work in both clean and noisy test environments except for the zero SNR level condition of certain noises.

This paper describes a study of a set of features based on cepstral mean and variance normalization (CMVN) plus auto regressive moving average (ARMA) filtering technique which is called MVA and on histogram equalization (HEQ) for robust speech recognition. First, we use MVA then HEQ in combination with CMVN and ARMA filtering as a post-processing module to mel frequency cepstral coefficients (MFCC), Relative Spectral-Perceptual linear prediction (RASTA-PLP) and power normalized cepstral coefficients (PNCC) features to improve the performance of the automatic speech recognition (ASR) system. The results on the Arabic database task have shown that both methods MVA and HEQ+ARMA improves the success rate for all features compared to the baseline system however HEQ was not found to perform better than MVA. The results also provide that RASTA-PLP outperforms PNCC and MFCC features.

An automatic speech recognition (ASR) is commonly used in these days. Current ASR systems perform well in ideal environment; however, it does not perform well in realistic noisy environment. As a robust ASR, ETSI has standardized Advanced Front-End (AFE) that adopts two-stage of iterative Wiener filter (IWF) to realize a speech enhancement as the front-end of ASR. In the ETSI AFE, 16 kHz speech is divided uniformly by QMF (Quadrature Mirror Filter) into lower-band and higher-band signal and only lower-band signal is emphasized by the IWF and MFCC (Mel Frequency Cepstral Coefficient) is extracted from both the second-stage of emphasized lower-band signal and non-emphasized higher-band signal. FFT is used to estimate speech spectrum that designs the Wiener filter. On the other hand, we have already proposed robust complex speech analysis for an analytic signal. It can estimate more robust and more accurate speech spectrum due to the introduced robust criterion and nature of analytic signal. This paper proposes an improved AFE using wide-band robust ELS (Extended Least Square) complex analysis and real-valued analysis instead of FFT. Moreover, the QMF synthesis is also introduced for second-stage of wide-band analysis. The experimental results using the CENSREC-2 speech database demonstrate that the performance is improved.

The communication mobile devices increasingly require automatic distributed speech recognition (ADSR) to be robust to the full range of real-environments noise and other acoustic distorting conditions. This paper aims to establish a solid framework integrating a noise-robust frontend in ADSR, which is lacking at present. The European Telecommunications Standards Institute advanced frontend (ETSI-AFE) is commonly referred method which performs very well in noisy environments, but poorly in noise-free conditions, and it is primarily suitable for dealing with stationary noise. We are intended to fill this gap and evaluate the development of an additional feature set for Mel frequency based ETSI-AFE, such as the power-normalized cepstral coefficients (PNCC). Thereafter, the stream weights of the resulting multi-stream hidden Markov models are optimized automatically by deploying a novel approach based on a discriminative model combination (DMC). This system DMC-MFCC frontend achieves significant performance improvements over related systems under a wide range of signal-to-noise ratios.KeywordsDistributed speech recognitionAdvanced frontendDiscriminative model combinationMFCCPNCC

ASR converts speech signal into corresponding text form. The performance of an ASR decreases under noisy environment. To overcome this problem a speech enhancement need to be performed on noisy speech before being fed to an ASR system. Speech enhancement techniques have been developed over past several decades, some of these techniques introduce musical noise. To achieve further improvement in recognition accuracy, a generalized distillation framework is used in which machines learns machines. In this paper, an ASR is implemented for noisy kannada language speech using generalized distillation framework. In this framework, a teacher machine is trained with clean speech and student machine with 4 different noise speech and teacher machine help student machine to learn by providing additional information needed. During test phase, a student machine is tested with 4 different noise speech other than used in training. A DNN acoustic model is build using a 39 dimension MFSC features and bi-gram language model is created using Kaldi Speech Recognition Toolkit. Experimental results shows that generalized distillation framework for kannada noisy speech achieved a reduction in WER compared to an HMM-GMM approach.

This work investigates the use of subglottal resonances (SGRs) for speaker normalization in noisy environments. Based on our previous work, a noise-robust algorithm is developed for estimating the first three SGRs from speech signals; it achieves robustness by factoring the short-term (or local) signal-to-noise ratio into the estimation process. The SGR estimates provided by this algorithm are refined by applying maximum-likelihood (ML) corrections, and are used in a non-linear frequency-warping technique that we recently developed. This SGR-based normalization (SN) scheme is evaluated on the AURORA-4 database in clean and noisy conditions. Using power-normalized cepstral coefficients (PNCCs) as front-end features, SN reduces the average word error rate by 8.7% relative to ML-based vocal-tract length normalization (VTLN). A fast version of SN (without ML corrections of SGR estimates) is also found to outperform VTLN (by 5.9% relative); it is computationally less complex than VTLN and hence a potential alternative for real-time applications.

The precedence effect describes the auditory system’s ability to suppress later-arriving components of sound in a reverberant environment, maintaining the perceived arrival azimuth of a sound in the direction of the actual source, even though the later reverberant components may arrive from other directions. It is also widely believed that precedence-like processing can also improve speech intelligibility for humans and the accuracy of speech recognition systems in reverberant environments. While the mechanisms underlying the precedence effect have traditionally been assumed to be binaural in nature, it is also possible that the suppression of later-arriving components may take place monaurally, and that the suppression of the corresponding components of the spatial image may be a consequence of this more peripheral processing. This paper compares potential contributions of onset enhancement (and consequent steady-state suppression) of the envelopes of subband components of speech at the monaural and binaural levels. Experimental results indicate that substantial improvement in recognition accuracy can be obtained in reverberant environments if feature extraction includes both onset enhancement and binaural interaction. Recognition accuracy appears to be relatively unaffected by which stage in the binaural processing is the site of the suppression mechanism. [Work supported by the LG Yonam Foundation and Cisco.]

Speech recognition is widely used in robot control and automation. Nevertheless, the use of speech recognition in robots is limited due to its susceptibility to background noise. This paper proposes a speech recognition algorithm to control robots in noisy environments. The proposed algorithm is based on Perceptual Linear Predictive Cepstral Coefficients (PNCC), which is a noise-resistant feature extraction technique, and Modified K-Nearest Neighbors (KNN) with Dynamic Time Warping (DTW) as the classifier. A new KNN-DTW classifier is proposed, integrating weighted KNN and DTW. The proposed algorithm results from experiments comparing PNCC and Mel-frequency cepstral coefficients (MFCC) feature extraction techniques with different classifiers, namely KNN-DTW, two types of KNN (weighted KNN and Medium-KNN), and two types of Support Vector Machine SVM (Linear SVM and Quadratic SVM). The database used to investigate the accuracy was the audio-visual data corpus database UOTletters, which includes 30 speakers, 26 English letters, and 1560 utterances. The database is divided into 50% for training and 50% for testing purposes. In a noise-free environment, the accuracy of the proposed algorithm reached 100%. Moreover, the proposed algorithm demonstrates greater noise immunity across all five noise levels, with an average accuracy difference of 13.67% compared to baseline algorithms.

In conventional Vector Taylor Series (VTS) based noisy speech recognition methods, Hidden Markov Models (HMMs) are trained using clean speech, and the parameters of the clean speech HMM are adapted to test noisy speech, or the original clean speech is estimated from the test noisy speech. However, these approaches have a drawback in that acoustic models trained using noisy speech cannot be used in recognition. In noisy speech recognition, improved performance is generally expected by employing noisy acoustic models produced by methods such as Multi-condition Training (MTR) and Multi-Model-based Speech Recognition framework (MMSR). Motivated by this idea, a method has been developed that can make use of the noisy acoustic models in the VTS algorithm where additive noise was adapted for the speech feature compensation. In this paper, we modified the previous method to adapt channel noise as well as additive noise. A mathematical relation was derived in the log-spectrum domain between the test and training noisy speech considering both channel and additive noise. After approximating the relation using VTS, Minimum Mean Square Error (MMSE) estimation of the training noisy speech is obtained from the test noisy speech based on the relation. The proposed method was applied to noisy speech HMMs trained by MTR and MMSR and cou ld reduce the relative word error rate by 7% and 8%, respectively, in the noisy speech recognition experiments on the Aurora 2 database.

CENSREC-4 evaluation framework has been distributed for evaluating distant-talking speech under various noisy and reverberant environments. It however has not been evaluated how variable noisy and reverberant features in this contains. We thus try to evaluate CENSREC-4 with our designed noisy and reverberant criteria based on PESQ and acoustic parameters. We specifically focus on criteria to represent the difficulty of noisy and reverberant speech recognition, and also confirm why it is difficult to easily evaluate the recognition accuracy in a part of CENSREC-4 corpus with our proposed noisy and reverberant criteria. We first designed the noisy and reverberant criteria using the relationship among the D value, the PESQ, and the ASR performance. We then tried to estimate the recognition accuracy in various noisy and reverberant environments with CENSREC-4. We carried out evaluation experiments to confirm the difficulty to easily evaluate the recognition accuracy in a part of CENSREC-4 courpus0. As a result of evaluation experiments, we confirmed that it was difficult to estimate the accuracy of noisy and reverberant speech recognition in heavy noisy and reverberant environment with CENSREC-4. We therefore confirmed that CENSREC-4 contained very challenging and variable noisy and reverberant data.

Typical hands-free voice capturing relies on recording a distant speaker using an array of microphones. When speech signal is recorded in a reverberant environment, the microphones capture speech corrupted by room reverberation. Increasing the distance between the speaker and recording microphones, the power ratio of the direct and late room reverberation signals is decreased, which in turn decreases the accuracy of automatic speech recognition. This paper presents a multichannel Wiener filter which exploits geometric information about early room reflection paths in order to reduce late reverberation at the filter output. As a result of such pre-processing, the accuracy of automatic speech recognition systems is improved. The presented word error rate (WER) results for simulated rooms with different reverberation times indicate that the proposed rake multichannel Wiener filtering provides an improvement in accuracy of automatic speech recognition from distance in reverberant environments. The gain in terms of word error rate achieved in considered acoustic conditions reaches several percent.