Space discriminative function for microphone array robust speech recognition

Based on W-disjoint orthogonality of speech mixtures, a space d,scnmlnative tunetlon was proposer1 to enumerate and localize competing speakers in the surrounding environments. Then, a Wiener-like postfiherer was developed to adaptively suppress interferences. Experimental results with a hands-free speech recognizer under various SNR and competing speakers settings show that nearly 69 % error reduction can be obtained with a two-channel small aperture microphone array against the conventional single microphone baseline system. Comparisons were made against traditional delay-and-sum and Griffiths-Jim adaptive beamforming techniques to further assess the effectiveness of this method.

Adaptive bands filter bank optimized by genetic algorithm for robust speech recognition system

Perceptual auditory filter banks such as Bark-scale filter bank are widely used as front-end processing in speech recognition systems.However,the problem of the design of optimized filter banks that provide higher accuracy in recognition tasks is still open.Owing to spectral analysis in feature extraction,an adaptive bands filter bank (ABFB) is presented.The design adopts flexible bandwidths and center frequencies for the frequency responses of the filters and utilizes genetic algorithm (GA) to optimize the design parameters.The optimization process is realized by combining the front-end filter bank with the back-end recognition network in the performance evaluation loop.The deployment of ABFB together with zero-crossing peak amplitude (ZCPA) feature as a front process for radial basis function (RBF) system shows significant improvement in robustness compared with the Bark-scale filter bank.In ABFB,several sub-bands are still more concentrated toward lower frequency but their exact locations are determined by the performance rather than the perceptual criteria.For the ease of optimization,only symmetrical bands are considered here,which still provide satisfactory results.

The Free-Flowing Voice and its Connection With the Mother Tongue

Does the native tongue confer greater authenticity and connection？ And how does this connect with languages acquired later in life？ From thirty years of directing, training, and auditioning actors from a range of ethnicities, I have believed that the mother-tongue has a particular and organic connection for an actor, one difficult to achieve in any other language. This belief was confounded in a laboratory conducted with Romanian actors, March 2013. The work was performed in both English and Romanian and it was with a sense of shock that I observed that the work was more vital, compelling, and physically and vocally engaged when they spoke in English. What were the factors at play here and what are the implications for future work？ Patsy Rodenburghas written of the giddy delight children find in language. Under what conditions does the native tongue evoke that ＂giddy delight＂ and where and when does it become an obstacle to such pleasure？

oncausal spatial prediction filtering based on an ARMA model

Conventional f-x prediction filtering methods are based on an autoregressive model. The error section is first computed as a source noise but is removed as additive noise to obtain the signal, which results in an assumption inconsistency before and after filtering. In this paper, an autoregressive, moving-average model is employed to avoid the model inconsistency. Based on the ARMA model, a noncasual prediction filter is computed and a self-deconvolved projection filter is used for estimating additive noise in order to suppress random noise. The 1-D ARMA model is also extended to the 2-D spatial domain, which is the basis for noncasual spatial prediction filtering for random noise attenuation on 3-D seismic data. Synthetic and field data processing indicate this method can suppress random noise more effectively and preserve the signal simultaneously and does much better than other conventional prediction filtering methods.

Hidden Markov Models for Automatic Speech Recognition

In this paper the authors look into the problem of Hidden Markov Models （HMM）： the evaluation, the decoding and the learning problem. The authors have explored an approach to increase the effectiveness of HMM in the speech recognition field. Although hidden Markov modeling has significantly improved the performance of current speech-recognition systems, the general problem of completely fluent speaker-independent speech recognition is still far from being solved. For example, there is no system which is capable of reliably recognizing unconstrained conversational speech. Also, there does not exist a good way to infer the language structure from a limited corpus of spoken sentences statistically. Therefore, the authors want to provide an overview of the theory of HMM, discuss the role of statistical methods, and point out a range of theoretical and practical issues that deserve attention and are necessary to understand so as to further advance research in the field of speech recognition.

The Joumey of the Magi： A Lyric Monologue for First and Second Voices and Three-in-One Character（s）

Although T. S. Eliot＇s ＂The Journey of the Magi＂ is a religious poem in the profoundest sense, the title of my paper is intended to give only a sly wink at Trinitarianism. My real object is to explain how Eliot contrived to manufacture a poem which, at fu＇st glance, resembles a dramatic monologue （generally understood as a poem for one voice----that of a historical/fictional/mythological character addressing a silent listener, group of listeners or reader）, yet which is slowly revealed as a lyrical monologue （for the poet＇s own voice） which yet--and this quite intentionally----contains considerably more than mere echoes of another two speakers： namely a Magus and the biblical translator and, most famously, sermon writer Archbishop Launcelot Andrewes （1555-1626） court preacher to James 1 and Charles 1 of England. I wish to show how Eliot, in writing what is ultimately confessional verse, goes out of his way to hoodwink the reader by allowing the first two of his ＂{The} Three Voices of Poetry＂ （1957） to overlap with and then incorporate the third. His own descriptions of these voices are （i） lyric, defined as ＂the poet talking to himself＂, （ii） that of the single speakerwho gives a （dramatic） monologuel ＂addressing an {imaginary} audience in an assumed voice＂ and （iii） that of the verse dramatist ＂who attempts to create a dramatic character speaking in verse when he {i.e. the author} is saying.., only what he can say within the limits of one imaginary character addressing another imaginary character＂ yet adding ＂some bit of himself that the author gives to a character may be the germ from which that character starts＂ （Eliot, 1957, pp. 38, 40）. The basis of my argument is that such an act of＂giving of the self＇ as the raw material for the creation of a dramatic monologue persona as well as a character designed for the stage had been part and parcel of Eliot＇s modus operandi up to and including ＂Prufrock＂ and The Waste Land; further, that in ＂The Journey of the Magi＂ and his later commentary upon it he fmally comes out and admits the fact, and in far clearer a manner than he does when defining the Objective Correlative in his essays on Hamlet. Far from attempting to erase the sense of selfhood from his poetry, I believe that Eliot, consciously or not, ended up by demonstrating to those who worshipped the Romantics and their cult of personality just how difficult it was to express the purely subjective self in poetry.

Transfer learning with deep sparse auto-encoder for speech emotion recognition

In order to improve the efficiency of speech emotion recognition across corpora,a speech emotion transfer learning method based on the deep sparse auto-encoder is proposed.The algorithm first reconstructs a small amount of data in the target domain by training the deep sparse auto-encoder,so that the encoder can learn the low-dimensional structural representation of the target domain data.Then,the source domain data and the target domain data are coded by the trained deep sparse auto-encoder to obtain the reconstruction data of the low-dimensional structural representation close to the target domain.Finally,a part of the reconstructed tagged target domain data is mixed with the reconstructed source domain data to jointly train the classifier.This part of the target domain data is used to guide the source domain data.Experiments on the CASIA,SoutheastLab corpus show that the model recognition rate after a small amount of data transferred reached 89.2%and 72.4%on the DNN.Compared to the training results of the complete original corpus,it only decreased by 2%in the CASIA corpus,and only 3.4%in the SoutheastLab corpus.Experiments show that the algorithm can achieve the effect of labeling all data in the extreme case that the data set has only a small amount of data tagged.

The Use of Children＇s Literature With University English as a Foreign Language Learners in Taiwan

This paper reports findings from a longitudinal qualitative study that investigated the use of children＇s literature for Taiwan Residents University English as a Foreign Language （EFL） students＇ reading. During the course of their sophomore year, 17 students participated and each student held two to seven individual reading sessions, to which they brought a self-selected children＇s picture storybook or children＇s novel they had finished reading on their own and orally read it to the researcher. Their oral reading and the discussion of each book with the researcher were audio recorded. To gain insight into the reading progress, these oral data were categorized and analyzed in terms of mispronunciation patterns, misunderstanding of vocabulary, misinterpretation of sentence or passage, and researcher＇s guidance. General findings of the 17 participants were presented in three categories： （1） vocabulary acquisition, （2） common comprehension problems, and （3） common pronunciation problems. Further analysis of two motivated students who read five to seven books revealed that （1） these two EFL learners gradually developed conscious awareness of their own pronunciation and comprehension errors and （2） they progressively acquired better competence to apply the pronunciation tips and reading comprehension techniques provided by the researcher during previous sessions. These findings and corresponding implications are discussed and further research suggestions are made.

Improved hidden Markov model for speech recognition and POS tagging

In order to overcome defects of the classical hidden Markov model (HMM), Markov family model (MFM), a new statistical model was proposed. Markov family model was applied to speech recognition and natural language processing. The speaker independently continuous speech recognition experiments and the part-of-speech tagging experiments show that Markov family model has higher performance than hidden Markov model. The precision is enhanced from 94.642% to 96.214% in the part-of-speech tagging experiments, and the work rate is reduced by 11.9% in the speech recognition experiments with respect to HMM baseline system.

HYBRID INTERFERENCE CANCELLATION SCHEME FOR BLIND MULTIUSER DETECTION IN MULTIPATH FADING CHANNELS

This paper introduces and analyzes a detection scheme for adaptive suppression of Multiuser Access Interference (MAI) and MultiPath Distortion (MPD) for mobile station of DS/CDMA system. The proposed detection scheme may amount to a RAKE receiver structure,wherein each branch is considered as a linear multiuser filter designed under a Linear Constrained Minimum Variance (LCMV) optimization strategy to suppress MAI, followed by a proper combining rule to suppress MPD. The adaptive blind multiuser detecting and optimum combining of the proposed receiver are realized, based on the Least-Mean-Square (LMS) algorithm and an adaptive vector tracking algorithm respectively. Finally, the feasibility of the above two algorithms is proved by the numerical results provided by computer simulation.

Sense of Harmony in Chinese Traditional Culture

A harmonious culture is not supposed to be a monologue of one voice but a dialogue of many voices. Chinese ancient culture doesn＇t mean one-dimensional notion of nature transforming to human like that in western philosophy, but means the hiding and dialogue between the nature and the human body. Specifically, Wen hua just means a dialogue based on the body. Heaven Unitfied With the Human and Complement of Confucianism and Taoism provide this opinion. Sense of Harmony in Chinese traditional culture provides us important ideological resources for building a harmonious culture and promoting the harmonious development of mankind.

The Great Semblance Is Invisible Lao tzu and Chuang tzu＇s Cognitive Aesthetics

In the paper I will research Lao tzu and Chuang tzu＇s cognitive aesthetics based on the generalized cognizance. Lao tzu and Chuang tzu are the representative figures of Taoism in pre-Qin period, they fully affirm the natural and human nature, naysay affirm human social, cultural and moral, cancel the traditional music, naysay affirm material, they pursue art which is eligibility natural and completely abandon the man-made things, esthetical state according with Tao. They think the great voice is no sound, the great semblance is invisible, the no sound voice and invisible semblance are insight native beauty which are associated with a specific aesthetic feeling but surpass the limited aesthetic feeling, all is the highest state of art and beauty, reach this level, in fact, has entered the Tao. In order to reach this state, they asked people to cleanse away inner desire and external disturbance, keep simplicity, abandon knowledge and wisdom, keep heart bright and clean, forget everything in order to contact the natural law with the nature, the nature＇s mystery runs automatically, the sounds of nature sound itself. About the invisible semblance aesthetic, they put forward the concepts such as gain its meaning but forget the word, illocutionary force

Research on the Accurate Computer Music Composition Mode under the Background of Big Data

In this paper, we conduct research on the accurate computer music composition mode under the basic background of big data. From the perspective of music theory, in terms of micro notes jump within a certain range, will not cause discomfort, notes there are upward at the same time, the downward makes the whole piece will not boring. From the macroscopic, the nature of the chord with internal connection between motivation and motivation as support, at the same time, the whole range can maintain in a certain range, making sound effects is can let a person feel cheerful. Computer science first occupied is music in the music sound production and the propagation of position as it is mainly used for digital audio clips and automated synthesis. But there is a certain distance away from the traditional concept of composition. Therefore, our research proposes the innovative perspective to deal with the issues for achieving the better performance.

Network Fault Analysis from Passive Measurement

A new passive method for automatic dis-covery and bcation of network failure is proposed. This method employs a passive measurement to collect infonmtion and events from network traffic, and em-ploys a rrodel-based reasoning system to detect and locate network faults. Measurement points are de-ployed in a backbone network to capture the traffic and then evaluate the Quality of Service （QoS） metrics of end-to-end IP conversations. A muting rrodel is al-so established for the observed network to simulate the attributes and activities of reuters and links. This muting model also deduces the muting path for each IP conversation, and thus the QoS metrics of IP con-versations are mapped into the metrics of paths. With the inforrmtion of shared links of overlapping paths and network torrography technique, the QoS metrics of links can also be estimated, and the poorly rated links are picked out as failure points. This method is imple-mented in a tool named FaultMan, which is deployed in a campus network. Test results have shown its availa-bility in rriddle-scale networks.

Tracing the Holistic Voice in Ecological Space： Exploring Theodore Roszak＇s Ecopsychology in Henri Lefebvre＇s ＂Differential Space＂

The goal of ecopsychology is to awaken the inherent sense of environmental reciprocity that lies within the ecological unconsciousness. Proclaiming the spirit of ecopsychology, Theodore Roszak argues that psychotherapy is an urban movement, but human beings can never heal themselves until they reconnect with nature. Other therapies aim at healing the alienation between person and person, person and family, person and society; ecopsychology intends to heal the more primary alienation between the person and the natural environment. Henri Lefebvre＇s work has revitalized urban studies, geography and planning via concepts like the social production of space. Lefebvre claims that space is not an inert, neutral, and pre-existing given, but rather, an on-going production of spatial relations. According to Lefebvre, space is produced by three types of practice： spatial practices of physical transformation of the environment, practices of representation of space, and everyday practices of representational space. Lefebvre further presents a ＂differential space,＂ named as such for its dialectical resistance to the forces of homogenization present in ＂abstract space.＂ The aim of this paper is to trace the ecological voice from Roszak＇s The Voice of the Earth in Henri Lefebvre＇s ＂differential space.＂ Roszak＇s ecopsychology has formed a differential space, acknowledging that the boundaries of dualism and separations such as mind and body, man and nature should be finally dissolved in terms of ecological sustainability. Within this space, a holistic approach and thinking are created and required to take into account perception of the inextricable relationship between all life and all phenomena.

从宋人笔记看“打”字音变

宋人笔记是一种口语化极强的文本,透过它可以看到,'打'字,不仅在宋代已经具有万能化倾向,其本身的字音问题也引起了当时文人的注意。欧阳修就这两方面都提出了自己的看法:……触事皆谓之'打'。而徧检字书,了无此字。丁雅反者其义主'考击'之'打'自音谪疑当作滴耿,以字学言之,'打'字从手、从丁,丁又击物之声,故音'谪耿'为是。

Past review,current progress,and challenges ahead on the cocktail party problem

The cocktail party problem,i.e.,tracing and recognizing the speech of a specific speaker when multiple speakers talk simultaneously,is one of the critical problems yet to be solved to enable the wide application of automatic speech recognition(ASR) systems.In this overview paper,we review the techniques proposed in the last two decades in attacking this problem.We focus our discussions on the speech separation problem given its central role in the cocktail party environment,and describe the conventional single-channel techniques such as computational auditory scene analysis(CASA),non-negative matrix factorization(NMF) and generative models,the conventional multi-channel techniques such as beamforming and multi-channel blind source separation,and the newly developed deep learning-based techniques,such as deep clustering(DPCL),the deep attractor network(DANet),and permutation invariant training(PIT).We also present techniques developed to improve ASR accuracy and speaker identification in the cocktail party environment.We argue effectively exploiting information in the microphone array,the acoustic training set,and the language itself using a more powerful model.Better optimization ob jective and techniques will be the approach to solving the cocktail party problem.

Numerical study on the correlation of transonic single-degree-of-freedom flutter and buffet

Transonic single-degree-of-freedom(SDOF) flutter and transonic buffet are the typical and complex aeroelastic phenomena in the transonic flow. In this study, transonic aeroelastic issues of an elastic airfoil are investigated using Unsteady Reynolds-Averaged Navier-Stokes(URANS) equations. The airfoil is free to vibrate in SDOF of pitching. It is found that, the coupling system may be unstable and SDOF self-excited pitching oscillations occur in pre-buffet flow condition, where the free-stream angle of attack(AOA) is lower than the buffet onset of a stationary airfoil. In the theory of classical aeroelasticity, this unstable phenomenon is defined as flutter. However, this transonic SDOF flutter is closely related to transonic buffet(unstable aerodynamic models) due to the following reasons. Firstly, the SDOF flutter occurs only when the free-stream AOA of the spring suspended airfoil is slightly lower than that of buffet onset, and the ratio of the structural characteristic frequency to the buffet frequency is within a limited range. Secondly, the response characteristics show a high correlation between the SDOF flutter and buffet. A similar "lock-in" phenomenon exists, when the coupling frequency follows the structural characteristic frequency. Finally, there is no sudden change of the response characteristics in the vicinity of buffet onset, that is, the curve of response amplitude with the free-stream AOA is nearly smooth. Therefore, transonic SDOF flutter is often interwoven with transonic buffet and shows some complex characteristics of response, which is different from the traditional flutter.

Self-powered acoustic source Iocator in underwater environment based on organic film triboelectric nano-generator

Detecting/sensing targets underwater has very important applications in environmental study, civil engineering and national security. In this paper, an organic-film based triboelectric nanogenerator （TENG） has been successfully demonstrated for the first time as a self-powered and high sensitivity acoustic sensor to detect underwater targets at low frequencies around 100 Hz. This innovative, cost-effective, simple-design TENG consists of a thin-film-based Cu electrode and a polytetrafluoroethylene （PTFE） film with nanostructures on its surfaces. On the basis of the coupling effect between triboelectrification and electrostatic induction, the sensor generates electrical output signals in response to incident sound waves. Operating at a resonance frequency of 110 Hz, under an acoustic pressure of 144.2 dBspc, the maximum open-circuit voltage and short-circuit current of the generator can respectively reach 65 V and 32 ~A underwater. The directional dependence pattern has a bi-directional shape with a total response angle of 60~. Its sensitivity is higher than -185 dB in the frequency range from 30 Hz to 200 Hz. The highest sensitivity is -146 dB at resonance frequency. The three-dimensional coordinates of an acoustic source were identified by four TENGs, self-powered active sensors, and the location of the acoustic source was determined with an error about 0.2 m. This study not only expands the application fields of TENGs from the atmosphere to water, but also shows the TENG is a promising acoustic source locator in underwater environments.

Effect of Plate Position and Size for Self-Induced Flow Oscillation of Underexpanded Supersonic Jet Impinging on Perpendicular Plate

When the underexpanded supersonic jet impinges on the obstacle, it is well known that the self-induced flow oscillation occurs at the specific condition of the pressure ratio in the flowfield, the position of an obstacle and so on. This oscillation is related with the noise problems of aeronautical and other industrial engineering so that the characteristic and the mechanism of self-induced flow oscillation have to be cleared to control the various noise problems. But, it seems that the characteristics of the oscillated flowfield and the mechanism of oscillation have to be more clear to control the oscillation. This paper aims to clarify the effect of the plate position and the width for the self-induced flow oscillation of an underexpanded supersonic jet impinging on the perpendicular plate by the experiment and the numerical analysis. From the results, it is clear that the occurring domain of the self-induced flow oscillation and its dimension strongly depend on the plate position and the width.