Exploiting foreign resources for DNN-based ASR

In this paper, the improvements in the recently implemented Kannada speech recognition system is demonstrated in detail. The Kannada automatic speech recognition (ASR) system consists of ASR models which are created by using Kaldi, IVRS call flow and weather and agricultural commodity prices information databases. The task specific speech data used in the recently developed spoken dialogue system had high level of different background noises. The different types of noises present in collected speech data had an adverse effect on the on line and off line speech recognition performances. Therefore, to improve the speech recognition accuracy in Kannada ASR system, a noise reduction algorithm is developed which is a fusion of spectral subtraction with voice activity detection (SS-VAD) and minimum mean square error spectrum power estimator based on zero crossing (MMSE-SPZC) estimator. The noise elimination algorithm is added in the system before the feature extraction part. An alternative ASR models are created using subspace Gaussian mixture models (SGMM) and deep neural network (DNN) modeling techniques. The experimental results show that, the fusion of noise elimination technique and SGMM/DNN based modeling gives a better relative improvement of 7.68% accuracy compared to the recently developed GMM-HMM based ASR system. The least word error rate (WER) acoustic models could be used in spoken dialogue system. The developed spoken query system is tested from Karnataka farmers under uncontrolled environment.

A deep neural network (DNN)-based speech enhancement (SE) aiming to maximize the performance of an automatic speech recognition (ASR) system is proposed in this paper. In order to optimize the DNN-based SE model in terms of the character error rate (CER), which is one of the metric to evaluate the ASR system and generally non-differentiable, our method uses two DNNs: one for speech processing and one for mimicking the output CERs derived through an acoustic model (AM). Then both of DNNs are alternately optimized in the training phase. Even if the AM is a black-box, e.g., like one provided by a third-party, the proposed method enables the DNN-based SE model to be optimized in terms of the CER since the DNN mimicking the AM is differentiable. Consequently, it becomes feasible to build CER-centric SE model that has no negative effect, e.g., additional calculation cost and changing network architecture, on the inference phase since our method is merely a training scheme for the existing DNN-based methods. Experimental results show that our method improved CER by 8.8% relative derived through a black-box AM although certain noise levels are kept.

The automatic speech recognition (ASR) system based on deep neural network is easy to be attacked by an adversarial example due to the vulnerability of neural network, which is a hot topic in recent years. The adversarial example does harm to the ASR system, especially if the common-dependent ASR goes wrong, it will lead to serious consequences. To improve the robustness and security of the ASR system, the defense method against adversarial examples must be proposed. Based on this idea, we propose an algorithm of devastation and detection on adversarial examples which can attack the current advanced ASR system. We choose advanced text-dependent and command-dependent ASR system as our target system. Generating adversarial examples by the OPT on text-dependent ASR and the GA-based algorithm on command-dependent ASR. The main idea of our method is input transformation of the adversarial examples. Different random intensities and kinds of noise are added to the adversarial examples to devastate the perturbation previously added to the normal examples. From the experimental results, the method performs well. For the devastation of examples, the original speech similarity before and after adding noise can reach 99.68%, the similarity of the adversarial examples can reach 0%, and the detection rate of the adversarial examples can reach 94%.

Combined speech enhancement and auditory modelling for robust distributed speech recognition

When an automatic speech recognition (ASR) system is deployed for real-world applications, it often receives only one utterance at a time for decoding. This single utterance could be of short duration depending on the ASR task. In these cases, robust estimation of speaker normalizing methods like feature-space maximum likelihood linear regression (FMLLR) and i -vectors may not be feasible. In this paper, we propose two unsupervised speaker normalization techniques—one at feature level and other at model level of acoustic modeling—to overcome the drawbacks of FMLLR and i -vectors in real-time scenarios. At feature level, we propose the use of deep neural networks (DNN) to generate pseudo-FMLLR features from time-synchronous pair of filterbank and FMLLR features. These pseudo-FMLLR features can then be used for DNN acoustic model training and decoding. At model level, we propose a generalized distillation framework, where a teacher DNN trained on FMLLR features guides the training and optimization of a student DNN trained on filterbank features. In both the proposed methods, the ambiguity in choosing the speaker-specific FMLLR transform can be reduced by augmenting i -vectors to the input filterbank features. Experiments conducted on 33-h and 110-h subsets of Switchboard corpus show that the proposed methods provide significant gains over DNNs trained on FMLLR, i -vector appended FMLLR, filterbank and i -vector appended filterbank features, in real-time scenario.

This paper describes a comparison between hybrid and end-to-end Automatic Speech Recognition (ASR) systems, which were evaluated on the IberSpeech-RTVE 2020 Speech-to-Text Transcription Challenge. Deep Neural Networks (DNNs) are becoming the most promising technology for ASR at present. In the last few years, traditional hybrid models have been evaluated and compared to other end-to-end ASR systems in terms of accuracy and efficiency. We contribute two different approaches: a hybrid ASR system based on a DNN-HMM and two state-of-the-art end-to-end ASR systems, based on Lattice-Free Maximum Mutual Information (LF-MMI). To address the high difficulty in the speech-to-text transcription of recordings with different speaking styles and acoustic conditions from TV studios to live recordings, data augmentation and Domain Adversarial Training (DAT) techniques were studied. Multi-condition data augmentation applied to our hybrid DNN-HMM demonstrated WER improvements in noisy scenarios (about 10% relatively). In contrast, the results obtained using an end-to-end PyChain-based ASR system were far from our expectations. Nevertheless, we found that when including DAT techniques, a relative WER improvement of 2.87% was obtained as compared to the PyChain-based system.

Conventional far-field automatic speech recognition (ASR) systems typically employ microphone array techniques for speech enhancement in order to improve robustness against noise or reverberation. However, such speech enhancement techniques do not always yield ASR accuracy improvement because the optimization criterion for speech enhancement is not directly relevant to the ASR objective. In this work, we develop new acoustic modeling techniques that optimize spatial filtering and long short-term memory (LSTM) layers from multi-channel (MC) input based on an ASR criterion directly. In contrast to conventional methods, we incorporate array processing knowledge into the acoustic model. Moreover, we initialize the network with beamformers' coefficients. We investigate effects of such MC neural networks through ASR experiments on the real-world far-field data where users are interacting with an ASR system in uncontrolled acoustic environments. We show that our MC acoustic model can reduce a word error rate (WER) by~16.5\% compared to a single channel ASR system with the traditional log-mel filter bank energy (LFBE) feature on average. Our result also shows that our network with the spatial filtering layer on two-channel input achieves a relative WER reduction of~9.5\% compared to conventional beamforming with seven microphones.

A statistically designed Automatic Speech Recognition (ASR) system extracts features from speech signals using feature extraction methods, links the extracted features with the expected phonetics of the hypothesis using acoustic models, uses language model to add prior information about the structure of the target language. For many years, Mel-frequency Cepstral coefficients (MFCC), n-gram, and Hidden Markov Model (HMM) approaches have been used predominantly for feature extraction, language modeling and acoustic modeling, respectively. However, performance degradation of MFCC in noisy conditions and inaccuracy of HMMs while handling large vocabularies have made researchers to propose more efficient methods. The proposed work uses noise robust method Gammatone Frequency Cepstral Coefficients(GFCC) for feature extraction, trigram language modeling, and HMM-Gaussian mixture model (GMM) based acoustic modeling to implement a continuous Hindi language ASR system. Also, it applies Differential Evolution (DE) technique to refine the GFCC features and discriminative techniques to enhance performance of the acoustic model. The performance of the implemented system has been evaluated by using different feature extraction methods, variants of n-gram language modeling techniques and different discriminative techniques in clean as well as noisy conditions. Initially, the results reveal that DE optimized GFCC with HMM-Gaussian Mixture Model (GMM) acoustic modeling performs better than MFCC, PLP and MF-PLP feature extraction methods. Secondly, the experimental results show that the Minimum Phone Error (MPE) outperforms Maximum Mutual Information (MMI) and Maximum Likelihood Estimation (MLE) and trigram based language modeling gives more accurate results than unigram and bigram language modeling. Finally, it has been concluded that the continuous Hindi language ASR system implemented using DE refined GFFC feature extraction method with MPE discriminative training technique and trigram based language modeling gives better accuracy in clean as-well-as noisy environments.

Some of the speech databases and large spoken language corpora that have been collected during the last fifteen years have been (at least partly) annotated with a broad phonetic transcription. Such phonetic transcriptions are often validated in terms of their resemblance to a handcrafted reference transcription. However, there are at least two methodological issues questioning this validation method. First, no reference transcription can fully represent the phonetic truth. This calls into question the status of such a transcription as a single reference for the quality of other phonetic transcriptions. Second, phonetic transcriptions are often generated to serve various purposes, none of which are considered when the transcriptions are compared to a reference transcription that was not made with the same purpose in mind. Since phonetic transcriptions are often used for the development of automatic speech recognition (ASR) systems, and since the relationship between ASR performance and a transcription’s resemblance to a reference transcription does not seem to be straightforward, we verified whether phonetic transcriptions that are to be used for ASR development can be justifiably validated in terms of their similarity to a purpose-independent reference transcription. To this end, we validated canonical representations and manually verified broad phonetic transcriptions of read speech and spontaneous telephone dialogues in terms of their resemblance to a handcrafted reference transcription on the one hand, and in terms of their suitability for ASR development on the other hand. Whereas the manually verified phonetic transcriptions resembled the reference transcription much closer than the canonical representations, the use of both transcription types yielded similar recognition results. The difference between the outcomes of the two validation methods has two implications. First, ASR developers can save themselves the effort of collecting expensive reference transcriptions in order to validate phonetic transcriptions of speech databases or spoken language corpora. Second, phonetic transcriptions should preferably be validated in terms of the application they will serve because a higher resemblance to a purpose-independent reference transcription is no guarantee for a transcription to be better suited for ASR development.

Standard automatic speech recognition (ASR) systems use acoustic models typically trained with speech of young adult speakers. Ageing is known to alter speech production in ways that require ASR systems to be adapted, in particular at the level of acoustic modeling. This paper reports ASR experiments that illustrate the impact of speaker age on speech recognition performance. A large read speech corpus in European Portuguese allowed us to measure statistically significant performance differences among age groups ranging from 60- to 90-year-old speakers. An increase of 41% relative (11.9% absolute) in word error rate was observed between 60-65-year-old and 81-86-year-old speakers. This paper also reports experiments on retraining acoustic models (AMs), further illustrating the impact of ageing on ASR performance. Differentiated gains were observed depending on the age range of the adaptation data use to retrain the acoustic models.

In this study, a real-time automatic speech recognition (ASR) system based on the Kaldi ASR toolkit, with low-latency and customizable models, without any internet connection, was developed. The proposed ASR system includes a voice activity detection (VAD) module and an audio transmitter as a front-end speech processing and a decoder for the received audio signals. The ASR system was evaluated in terms of ASR accuracy and speech processing speed. Consequently, the ASR system achieved high ASR accuracy on the CSJ (Corpus of Spontaneous Japanese) test set with super low-latency.

We report insights from translating Spanish conversational telephone speech into English text by cascading an automatic speech recognition (ASR) system with a statistical machine translation (SMT) system. The key new insight is that the informal register of conversational speech is a greater challenge for ASR than for SMT: the BLEU score for translating the reference transcript is 64%, but drops to 32% for translating automatic transcripts, whose word error rate (WER) is 40%. Several strategies are examined to mitigate the impact of ASR errors on the SMT output: (i) providing the ASR lattice, instead of the 1-best output, as input to the SMT system, (ii) training the SMT system on Spanish ASR output paired with English text, instead of Spanish reference transcripts, and (iii) improving the core ASR system. Each leads to consistent and complementary improvements in the SMT output. Compared to translating the 1-best output of an ASR system with 40% WER using an SMT system trained on Spanish reference transcripts, translating the output lattice of a better ASR system with 35% WER using an SMT system trained on ASR output improves BLEU from 32% to 38%.

This paper describes open source software (available at https://github.com/robotology/natural-speech) to build automatic speech recognition (ASR) systems and run them within the YARP platform. The toolkit is designed (i) to allow non-ASR experts to easily create their own ASR system and run it on iCub and (ii) to build deep learning-based models specifically addressing the main challenges an ASR system faces in the context of verbal human–iCub interactions. The toolkit mostly consists of Python, C++ code and shell scripts integrated in YARP. As additional contribution, a second codebase (written in Matlab) is provided for more expert ASR users who want to experiment with bio-inspired and developmental learning-inspired ASR systems. Specifically, we provide code for two distinct kinds of speech recognition: “articulatory” and “unsupervised” speech recognition. The first is largely inspired by influential neurobiological theories of speech perception which assume speech perception to be mediated by brain motor cortex activities. Our articulatory systems have been shown to outperform strong deep learning-based baselines. The second type of recognition systems, the “unsupervised” systems, do not use any supervised information (contrary to most ASR systems, including our articulatory systems). To some extent, they mimic an infant who has to discover the basic speech units of a language by herself. In addition, we provide resources consisting of pre-trained deep learning models for ASR, and a 2.5-h speech dataset of spoken commands, the VoCub dataset, which can be used to adapt an ASR system to the typical acoustic environments in which iCub operates.

This paper presents an analysis-by-synthesis approach for acoustic model adaptation. Using artificial speech data for speech recognition systems adaptation, has the potential to address the problem of data sparseness, to avoid speech recordings in real conditions and to provide the capability of performing large number of development cycles for Automatic Speech Recognition (ASR) systems in shorter time. The proposed adaptation framework uses unified ASR and synthesis system to produce artificial adaptation speech signals. In order to confirm the usability of the proposed approach, several experiments were performed where the artificial speech data was coded-decoded by different speech and waveform coders and the acoustic model used for synthesis was adapted for each coder. The recognition results show that the proposed method could be used successfully in the process of speech recognition systems performance assessment and improvement, not only for coded speech effects evaluation and adaptation, but also for other environment conditions.

The paper presents a system for speech recognition and synthesis for the Kazakh and Russian languages. It is designed for use by speakers of Kazakh; due to the prevalence of bilingualism among Kazakh speakers, it was considered essential to design a bilingual Kazakh-Russian system. Developing our system involved building a text processing and transcription system that deals with both Kazakh and Russian text, and is used in both speech synthesis and recognition applications. We created a Kazakh TTS voice and an additional Russian voice using the recordings of the same bilingual voice artist. A Kazakh speech database was collected and used to train deep neural network acoustic models for the speech recognition system. The resulting models demonstrated sufficient performance for practical applications in interactive voice response and keyword spotting scenarios.