This paper presents an electrolaryngeal (EL) speech enhancement method capable of significantly improving naturalness of EL speech while causing no degradation in its intelligibility. An electrolarynx is an external device that artificially generates excitation sounds to enable laryngectomees to produce EL speech. Although proficient laryngectomees can produce quite intelligible EL speech, it sounds very unnatural due to the mechanical excitation produced by the device. Moreover, the excitation sounds produced by the device often leak outside, adding to EL speech as noise. To address these issues, there are mainly two conventional approached to EL speech enhancement through either noise reduction or statistical voice conversion (VC). The former approach usually causes no degradation in intelligibility but yields only small improvements in naturalness as the mechanical excitation sounds remain essentially unchanged. On the other hand, the latter approach significantly improves naturalness of EL speech using spectral and excitation parameters of natural voices converted from acoustic parameters of EL speech, but it usually causes degradation in intelligibility owing to errors in conversion. We propose a hybrid approach using a noise reduction method for enhancing spectral parameters and statistical voice conversion method for predicting excitation parameters. Moreover, we further modify the prediction process of the excitation parameters to improve its prediction accuracy and reduce adverse effects caused by unvoiced/voiced prediction errors. The experimental results demonstrate the proposed method yields significant improvements in naturalness compared with EL speech while keeping intelligibility high enough.

Non-verbal communication incorporating visual, audio, and contextual information is important to make sense of and navigate the social world. Individuals who have trouble with social situations often have difficulty recognizing these sorts of non-verbal social signals. In this article, we propose a training tool NOCOA+ (Non-verbal COmmuniation for Autism plus) that uses utterances in visual and audio modalities in non-verbal communication training. We describe the design of NOCOA+, and further perform an experimental evaluation in which we examine its potential as a tool for computer-based training of non-verbal communication skills for people with social and communication difficulties. In a series of four experiments, we investigated 1) the effect of temporal context on the ability to recognize social signals in testing context, 2) the effect of modality of presentation of social stimulus on ability to recognize non-verbal information, 3) the correlation between autistic traits as measured by the autism spectrum quotient (AQ) and non-verbal behavior recognition skills measured by NOCOA+, 4) the effectiveness of computer-based training in improving social skills. We found that context information was helpful for recognizing non-verbal behaviors, and the effect of modality was different. The results also showed a significant relationship between the AQ communication and socialization scores and non-verbal communication skills, and that social skills were significantly improved through computer-based training.

Real-time machine speech translation systems mimic human interpreters and translate incoming speech from a source language to the target language in real-time. Such systems can be achieved by performing low-latency processing in ASR (automatic speech recognition) module before passing the output to MT (machine translation) and TTS (text-to-speech synthesis) modules. Although several studies recently proposed sequence mechanisms for neural incremental ASR (ISR), these frameworks have a more complicated training mechanism than the standard attention-based ASR because they have to decide the incremental step and learn the alignment between speech and text. In this paper, we propose attention-transfer ISR (AT-ISR) that learns the knowledge from attention-based non-incremental ASR for a low delay end-to-end speech recognition. ISR comes with a trade-off between delay and performance, so we investigate how to reduce AT-ISR delay without a significant performance drop. Our experiment shows that AT-ISR achieves a comparable performance to the non-incremental ASR when the incremental recognition begins after the speech utterance reaches 25% of the complete utterance length. Additional experiments to investigate the effect of ISR on translation tasks are also performed. The focus is to find the optimum granularity of the output unit. The results reveal that our end-to-end subword-level ISR resulted in the best translation quality with the lowest WER and the lowest uncovered-word rate.

This study proposes a variable selection linear regression (VSLR) adaptation framework to improve the accuracy of automatic speech recognition (ASR) with only limited and unlabeled adaptation data. The proposed framework can be divided into three phases. The first phase prepares multiple variable subsets by applying a ranking filter to the original regression variable set. The second phase determines the best variable subset based on a pre-determined performance evaluation criterion and computes a linear regression (LR) mapping function based on the determined subset. The third phase performs adaptation in either model or feature spaces. The three phases can select the optimal components and remove redundancies in the LR mapping function effectively and thus enable VSLR to provide satisfactory adaptation performance even with a very limited number of adaptation statistics. We formulate model space VSLR and feature space VSLR by integrating the VS techniques into the conventional LR adaptation systems. Experimental results on the Aurora-4 task show that model space VSLR and feature space VSLR, respectively, outperform standard maximum likelihood linear regression (MLLR) and feature space MLLR (fMLLR) and their extensions, with notable word error rate (WER) reductions in a per-utterance unsupervised adaptation manner.

An efficient, scalable speech recognition architecture combining topic detection and topic-dependent language modeling is proposed for multi-domain spoken language systems. In the proposed approach, the inferred topic is automatically detected from the user's utterance, and speech recognition is then performed by applying an appropriate topic-dependent language model. This approach enables users to freely switch between domains while maintaining high recognition accuracy. As topic detection is performed on a single utterance, detection errors may occur and propagate through the system. To improve robustness, a hierarchical back-off mechanism is introduced where detailed topic models are applied when topic detection is confident and wider models that cover multiple topics are applied in cases of uncertainty. The performance of the proposed architecture is evaluated when combined with two topic detection methods: unigram likelihood and SVMs (Support Vector Machines). On the ATR Basic Travel Expression Corpus, both methods provide a significant reduction in WER (9.7% and 10.3%, respectively) compared to a single language model system. Furthermore, recognition accuracy is comparable to performing decoding with all topic-dependent models in parallel, while the required computational cost is much reduced.

This paper introduces an evaluation framework for Japanese noisy speech recognition named AURORA-2J. Speech recognition systems must still be improved to be robust to noisy environments, but this improvement requires development of the standard evaluation corpus and assessment technologies. Recently, the Aurora 2, 3 and 4 corpora and their evaluation scenarios have had significant impact on noisy speech recognition research. The AURORA-2J is a Japanese connected digits corpus and its evaluation scripts are designed in the same way as Aurora 2 with the help of European Telecommunications Standards Institute (ETSI) AURORA group. This paper describes the data collection, baseline scripts, and its baseline performance. We also propose a new performance analysis method that considers differences in recognition performance among speakers. This method is based on the word accuracy per speaker, revealing the degree of the individual difference of the recognition performance. We also propose categorization of modifications, applied to the original HTK baseline system, which helps in comparing the systems and in recognizing technologies that improve the performance best within the same category.

In this paper, we propose a placement and routing method for a new memory-based programmable logic device (MPLD) and confirm its capability by placing and routing benchmark circuits. An MPLD consists of multiple-output look-up tables (MLUTs) that can be used as logic and/or routing elements, whereas field programmable gate arrays (FPGAs) consist of LUTs (logic elements) and switch blocks (routing elements). MPLDs contain logic circuits more efficiently than FPGAs because of their flexibility and area efficiency. However, directly applying the existing placement and routing algorithms of FPGAs to MPLDs overcrowds the placed logic cells and causes a shortage of routing domains between logic cells. Our simulated annealing-based method considers the detailed wire congestion and nearness between logic cells based on the cost function and reserves the area for routing. In the experiments, our method reduced wire congestion and successfully placed and routed 27 out of 31 circuits, 13 of which could not be placed or routed using the versatile place and route tool (VPR), a well-known method for FPGAs.

An accurate audio-visual speech corpus is inevitable for talking-heads research. This paper presents our audio-visual speech corpus collection and proposes a head-movement normalization method and a facial motion generation method. The audio-visual corpus contains speech data, movie data on faces, and positions and movements of facial organs. The corpus consists of Japanese phoneme-balanced sentences uttered by a female native speaker. An accurate facial capture is realized by using an optical motion-capture system. We captured high-resolution 3D data by arranging many markers on the speaker's face. In addition, we propose a method of acquiring the facial movements and removing head movements by using affine transformation for computing displacements of pure facial organs. Finally, in order to easily create facial animation from this motion data, we propose a technique assigning the captured data to the facial polygon model. Evaluation results demonstrate the effectiveness of the proposed facial motion generation method and show the relationship between the number of markers and errors.

The contribution of this paper is two-fold. Firstly, we conduct a large-scale real-world evaluation of the effectiveness of integrating an automatic transliteration system with a machine translation system. A human evaluation is usually preferable to an automatic evaluation, and in the case of this evaluation especially so, since the common machine translation evaluation methods are affected by the length of the translations they are evaluating, often being biassed towards translations in terms of their length rather than the information they convey. We evaluate our transliteration system on data collected in field experiments conducted all over Japan. Our results conclusively show that using a transliteration system can improve machine translation quality when translating unknown words. Our second contribution is to propose a novel Bayesian model for unsupervised bilingual character sequence segmentation of corpora for transliteration. The system is based on a Dirichlet process model trained using Bayesian inference through blocked Gibbs sampling implemented using an efficient forward filtering/backward sampling dynamic programming algorithm. The Bayesian approach is able to overcome the overfitting problem inherent in maximum likelihood training. We demonstrate the effectiveness of our Bayesian segmentation by using it to build a translation model for a phrase-based statistical machine translation (SMT) system trained to perform transliteration by monotonic transduction from character sequence to character sequence. The Bayesian segmentation was used to construct a phrase-table and we compared the quality of this phrase-table to one generated in the usual manner by the state-of-the-art GIZA++ word alignment process used in combination with phrase extraction heuristics from the MOSES statistical machine translation system, by using both to perform transliteration generation within an identical framework. In our experiments on English-Japanese data from the NEWS2010 transliteration generation shared task, we used our technique to bilingually co-segment the training corpus. We then derived a phrase-table from the segmentation from the sample at the final iteration of the training procedure, and the resulting phrase-table was used to directly substitute for the phrase-table extracted by using GIZA++/MOSES. The phrase-table resulting from our Bayesian segmentation model was approximately 30% smaller than that produced by the SMT system's training procedure, and gave an increase in transliteration quality measured in terms of both word accuracy and F-score.

Chinese is a representative tonal language, and it has been an attractive topic of how to process tone information in the state-of-the-art large vocabulary speech recognition system. This paper presents a novel way to derive an efficient phoneme set of tone-dependent units to build a recognition system, by iteratively merging a pair of tone-dependent units according to the principle of minimal loss of the Mutual Information (MI). The mutual information is measured between the word tokens and their phoneme transcriptions in a training text corpus, based on the system lexical and language model. The approach has a capability to keep discriminative tonal (and phoneme) contrasts that are most helpful for disambiguating homophone words due to lack of tones, and merge those tonal (and phoneme) contrasts that are not important for word disambiguation for the recognition task. This enables a flexible selection of phoneme set according to a balance between the MI information amount and the number of phonemes. We applied the method to traditional phoneme set of Initial/Finals, and derived several phoneme sets with different number of units. Speech recognition experiments using the derived sets showed its effectiveness.

Given a question and a set of its candidate answers, the task of answer validation (AV) aims to return a Boolean value indicating whether a given candidate answer is the correct answer to the question. Unlike previous works, this paper presents an unsupervised model, called the U-model, for AV. This approach regards AV as a classification task and investigates how effectively using redundancy of the Web into the proposed architecture. Experimental results with TREC factoid test sets and Chinese test sets indicate that the proposed U-model with redundancy information is very effective for AV. For example, the top@1/mrr@5 scores on the TREC05, and 06 tracks are 40.1/51.5% and 35.8/47.3%, respectively. Furthermore, a cross-model comparison experiment demonstrates that the U-model is the best among the redundancy-based models considered. Even compared with a syntax-based approach, a supervised machine learning approach and a pattern-based approach, the U-model performs much better.

This paper proposes a new speaker adaptation method using a speaker weighting technique for multiple reference speaker training of a hidden Markov model (HMM). The proposed method considers the similarities between an input speaker and multiple reference speakers, and use the similarities to control the influence of the reference speakers upon HMM. The evaluation experiments were carried out through the/b, d, g, m, n, N/phoneme recognition task using 8 speakers. Average recognition rates were 68.0%, 66.4%, and 65.6% respectively for three test sets which have different speech styles. These were 4.8%, 8.8%, and 10.5% higher than the rates of the spectrum mapping method, and also 1.6%, 6.7%, and 8.2% higher than the rates of the multiple reference speaker training, the supplemented HMM. The evaluation experiments clarified the effectiveness of the proposed method.

Although data compression is popularly used, compressed data have a problem that they are very sensitive to errors. This paper proposes a single burst error recovery method for Huffman coding by using the bidirectionally decodable Huffman coding. Computer simulation shows that the proposed method can recover 2.5lburst bits burst error with high probability, where lburst is the maximum length of burst errors which the proposed method is expected to be able to recover.

This paper discusses "face-to-talk" audio-visual speech detection for robust speech recognition in noisy environment, which consists of facial orientation based switch and audio-visual speech section detection. Most of today's speech recognition systems must actually turned on and off by a switch e.g. "push-to-talk" to indicate which utterance should be recognized, and a specific speech section must be detected prior to any further analysis. To improve usability and performance, we have researched how to extract the useful information from visual modality. We implemented a facial orientation based switch, which activates the speech recognition during a speaker is facing to the camera. Then, the speech section is detected by analyzing the image of the face. Visual speech detection is robust to audio noise, but because the articulation starts prior to the speech and lasts longer than the speech, the detected section tends to be longer and ends up with insertion errors. Therefore, we have fused the audio-visual modality detected sections. Our experiment confirms that the proposed audio-visual speech detection method improves recognition performance in noisy environment.

In current HMM based speech recognition systems, it is difficult to supplement acoustic spectrum features with additional information such as pitch, gender, articulator positions, etc. On the other hand, Bayesian Networks (BN) allow for easy combination of different continuous as well as discrete features by exploring conditional dependencies between them. However, the lack of efficient algorithms has limited their application in continuous speech recognition. In this paper we propose new acoustic model, where HMM are used for modeling of temporal speech characteristics and state probability model is represented by BN. In our experimental system based on HMM/BN model, in addition to speech observation variable, state BN has two more (hidden) variables representing noise type and SNR value. Evaluation results on AURORA2 database showed 36.4% word error rate reduction for closed noise test which is comparable with other much more complex systems utilizing effective adaptation and noise robust methods.

Brain-computer interfaces (BCIs) have been used by users to convey their intentions directly with brain signals. For example, a spelling system that uses EEGs allows letters on a display to be selected. In comparison, previous studies have investigated decoding speech information such as syllables, words from single-trial brain signals during speech comprehension, or articulatory imagination. Such decoding realizes speech recognition with a relatively short time-lag and without relying on a display. Previous magnetoencephalogram (MEG) research showed that a template matching method could be used to classify three English sentences by using phase patterns in theta oscillations. This method is based on the synchronization between speech rhythms and neural oscillations during speech processing, that is, theta oscillations synchronized with syllabic rhythms and low-gamma oscillations with phonemic rhythms. The present study aimed to approximate this classification method to a BCI application. To this end, (1) we investigated the performance of the EEG-based classification of three Japanese sentences and (2) evaluated the generalizability of our models to other different users. For the purpose of improving accuracy, (3) we investigated the performances of four classifiers: template matching (baseline), logistic regression, support vector machine, and random forest. In addition, (4) we propose using novel features including phase patterns in a higher frequency range. Our proposed features were constructed in order to capture synchronization in a low-gamma band, that is, (i) phases in EEG oscillations in the range of 2-50 Hz from all electrodes used for measuring EEG data (all) and (ii) phases selected on the basis of feature importance (selected). The classification results showed that, except for random forest, most classifiers perform similarly. Our proposed features improved the classification accuracy with statistical significance compared with a baseline feature, which is a phase pattern in neural oscillations in the range of 4-8 Hz from the right hemisphere. The best mean accuracy across folds was 55.9% using template matching trained by all features. We concluded that the use of phase information in a higher frequency band improves the performance of EEG-based sentence classification and that this model is applicable to other different users.