Speech recognition based on a combination of acousticfeatures with articulatory information

The contributions of the static and dynamic articulatory information to speech recognition were evaluated, and the recognition approaches by combining the articulatory information with acoustic features were discussed. Articulatory movements were observed by the Electromagnetic Articulographic System for reading speech, and the speech signals were recorded simultaneously. First, we conducted several speech recognition experiments by using articulatory features alone, consisting of a number of specific articulatory channels, to evaluate the contribution of each observation point on articulators. Then, the displacement information of articulatory data were combined with acoustic features directly and adopted in speech recognition. The results show that articulatory information provides with additional information for speech recognition which is not encoded in acoustic features. Furthermore, the contribution of the dynamic information of the articulatory data was evaluated by combining them in speech recognition. It is found that the second derivative of articulatory information provided quite larger contribution to speech recognition comparing with the second derivative of acoustical information. At last, the combination methods of articulatory features and acoustic ones were investigated for speech recognition. The basic approach is that the Bayesian Network (BN) is added to each state of HMM, where the articulatory information is represented by the BN as a factor of observed signals during training the model and is marginalized as a hidden variable in recognition stage. Results based on this HMM/BN framework show a better performance than the traditional method.     

Elimination of Spiral Waves and Competition between Travelling Wave Impulses and Spiral Waves

The interaction between travelling wave impulses and spiral waves is studied and the results of their competition are related to the exciting period. From the results, it is known that the formation and development of spiral waves in cardiac tissue depend on the period by which the travelling wave impulses are excited. A method is proposed to eliminate spiral waves, which is easily operated.     

第六讲水下声学传感器网络的发展和应用

文章主要介绍了正在发展中的水下声学传感器网络的一些概念和发展现状。在回顾了水声通信和水下网络的发展历史之后，首先从网络拓扑结构出发，指出多跳对等网络是适合于水声应用的一种网络结构，接着根据网络协议栈的概念，说明水下声学传感器网络在各协议层上所涉及的问题，并从物理层面，分析了水声介质的特殊性，和水下声学传感器网络所面临的一些主要的技术挑战，介绍了组成水下传感器网络的关键部件。以美国海网（Seaweb）为例，给出了一个水声传感器网实现的实例，对它的年度实验进行了介绍，最后，给出了中国未来的发展应用前景。     

Advances in micro-perforated panel absorbers

Theoretical and experimental investigations on the performance of micro-perforated -panel absorbers are reviewed in this paper. By reviewing recent research work, this paper reveals a relationship between the maximum absorption coefficient and the limit of the absorption frequency bandwidth. It has been demonstrated that the absorption frequency bandwidth can be extended up to 3 or 4 octaves as the diameters of the micro-holes decrease. This has become possible with the development of the technologies for manufacturing micro-perforated panels, such as laser drilling, powder metallurgy, welded meshing and electro-etching to form micrometer order holes. In this paper, absorption characteristics of such absorbers in random fields and in high sound intensity are discussed both theoretically and experimentally. A new absorbing structure based on micro-perforated-panel absorbers demonstrate experimentally high sound absorption capability. This review shows that the micro-perforated-panel absorber has potentials to be one of ideal absorbing materials in the 21st century.     

Statistical survey of female monophthong formants in Chinese Putonghua

In order to investigate the group characteristics of Putonghua monophthong for-mants, the tokens of 90 female students were surveyed. The formants were measured using LPC method. The averaged values and spread of formant frequencies were given with statistical meaning. The results show the difference from the previous measurements by other researchers decades ago. For all monophthongs, F4/F3 and F5/F4 are generally around 1.4. To discriminate monophthongs, F2/F1 and F3/F2 are possibly the two new parameters besides the first three formants.     

Acoustic vector sensor signal processing

Acoustic vector sensor simultaneously, colocately and directly measures orthogonal components of particle velocity as well as pressure at single point in acoustic field so that is possible to improve performance of traditional underwater acoustic measurement devices or detection systems and extends new ideas for solving practical underwater acoustic engineering problems. Although acoustic vector sensor history of appearing in underwater acoustic area is no long, but with huge and potential military demands, acoustic vector sensor has strong development trend in last decade, it is evolving into a one of important underwater acoustic technology. Under this background, we try to review recent progress in study on acoustic vector sensor signal processing, such as signal detection, DOA estimation, beamforming, and so on.     

The acoustic characteristics of clause, sentence and paragraph boundaries in Chinese Putonghua

The acoustic and perceptive characteristics of discourse units, including clauses, simple sentences, compound sentences and paragraphs, were studied. The acoustic characteristics were based on the analysis of a speech corpus consisted of ten monologues. The Perceived Boundary Strength (PBS) of the monologues was labeled on a five-point scale by 25 participants. The main results are as follows: (1) There is a significant difference in the PBS of clauses (corresponding to intonational phrase), sentences (including simple sentences and compound sentences) and paragraphs. But the PBS of simple sentences and compound sentences is the same, and the pitch reset and pause are also not significantly different between these two kinds of boundaries. (2) The pitch-based cue used to distinguish clauses, sentences and paragraphs appears to be the difference between pre- and post-boundary syllables, but not simply the pitch value of either syllable alone. (3) Declination mainly occurs within intonational phrases, whereas there is no consistent global declination in sentences or in paragraphs. (4) Pauses are longer and more varied at the boundary of larger discourse units. Moreover, there is a significant positive correlation between pause duration and degree of pitch reset at clause boundaries, but not at sentence or paragraph boundaries.     

Stochastic Resonance in an Over-Damped Bias Linear System with Dichotomous Noise

The stochastic resonance in an over-damped bias linear system subject to multiplicative and additive dichotomous noise （DN） is investigated. By using the linear-response theory and the properties of the DN, the exact expressions are found for the signal-to-noise ratio （SNR）. It is shown that the SNR is a non-monotonic function of the correlation time of the additive DN, and it varies non-monotonically with the bias of the external field, the intensity and asymmetry of the multiplicative DN, as well as the external field frequency. Moreover, the SNR depends on the bias of the system, as well as the strength and asymmetry of the additive DN.     

Research on relationships between Lamb wave velocity and static stress in metal plate

On the fact that an isotropic metal solid produces anisotropic property in the state of static stress, based on the theory of the nonlinear acoustoelasticity, the equivalent second order elastic constants are calculated for metal plate with static stress. For the case of thin metal plate with stress, the two kinds of dispersion equation for Lamb waves propagating parallel and vertical to the direction of static stress are derived. Using the equations, the relationships between Lamb wave velocity and static stress in a metal plate are discussed.     

Practical absorption limits of MPP absorber

The construction and properties of microperforated panel (MPP) absorber are discussed. The absorption limit of the absorber had been shown that low values of the perforate constant k = d(f/10)1/2 and the orifice diameter d (in mm) are essential for MPP to have high absorption in wide frequency band. To find the exact limits, take 1 for k as a start, because both specific resistance and high absorption require k around one. And the orifice diameter d is chosen as 0.1 mm, so that the peak absorption coefficient (resonance absorption) is at 1000 Hz, and high sound frequency may be in the absorption region. Is it possible for a single layer of such an MPP to cover the whole absorption region required in practice? The half-absorption limit is not a good criterion, because low absorption comes in also in some cases. The 0.5 absorption coefficient limit is suggested for practical region, as a standard for comparison. Absorption curves were drawn for different load resistances, of absorption coefficients versus frequency. Ordinary MPP absorber absorbs in slightly over two octaves, and the new absorber with r = 1 (specific resistance equal to the characteristic impedance in air)is slightly better than these, 2.5 octaves. The new absorbers with r > 1, are much better than these, and some satisfies high absorption in broad frequency range. Realization of these will mean great progress of MPP absorbers.