15 February 2017, Volume 57 Issue 2

    

• Select all
  |

MECHANICAL ENGINEERING
• Select
  Multi-axis coupling of machining centers based on electromechanical modeling

  ZHAO Tong, XIE Yi, HU Jian

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 113-119. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.001

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  The dynamic accuracy of a machining center is strongly influenced by its structural dynamics and control loops. The frequency response of the entire machine including the multi-directional dynamics is important in the machine tool design phase. The structural dynamics of a twin ball screw feed machine are analyzed using a finite element analysis (FEA) and measurements with a focus on the multi-axis coupling of its mechanical system. Then, an electromechanical modeling method is developed where the structural dynamics are replaced by transfer functions to study the multi-axis coupling of the electromechanical system of a horizontal machining center. The results show that when there are modal shapes leading to displacements of the tool or the workpiece in multiple directions, the multi-axis coupling of the machining center must be considered in studies of the dynamics and control parameters.
  
• Select
  B axis rotating machining of microstructured optical molds with rotational symmetry based on ultra-precision cutting

  GAO Xing, LI Yong, ZHONG Hao, YUE Quan, LI Chaojiang

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 120-127. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.002

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  A B axis rotating machining process based on ultra-precision cutting was developed to prevent burrs, sharp corner defects and surface scratches when machining microstructured optical molds with rotational symmetry. The form accuracy and surface quality of the molds were mainly influenced by the machining path, cutting speed, B axis rotating speed and cutting depth. Tests of these factors were used to optimize the cutting trajectory planning and machining parameters. Two microstructured molds were constructed with no sharp corner defects and no surface scratches, dimensional accuracies of 1.4 μm and surface roughnesses better than 14 nm. The results show the feasibility of the B axis rotating machining process for cutting microstructured optical molds with rotational symmetry.
  
• Select
  Digital workshop information model and its standardization

  ZHANG Zhaokun, SHAO Zhufeng, WANG Liping, ZHAO Qinzhi, ZHANG Yunfeng

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 128-133,140. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.003

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Digital workshops are key to intelligent manufacturing and intelligent plants. A standardized information model is needed to eliminate information silos and to achieve interoperability between workshops and equipment. This paper identifies the connotation, elements and requirements of a good digital workshop information model. The characteristics of existing information modeling methods are searched to find which modeling method is best for intelligent manufacturing digital workshops. Then, different information standardization methods are analyzed to develop guidelines for future information model standardization.
AEROSPACE ENGINEERING
• Select
  Compound action potential and personalized parameter regulation of vagus nerve stimulation

  YUAN Yuan, HAO Hongwei, LI Luming

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 134-140. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.004

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Vagus nerve stimulation (VNS) is widely used to treat various refractory diseases, such as epilepsy and depression. Personalized treatment is very important in these clinical applications. The intraoperative compound action potential (CAP) response of the vagus nerve is analyzed here to evaluate the personalized parameter regulation. A CAP detection system with in vivo experiments with pigs is used to analyze the relationship between the CAP and the VNS parameters, including the current amplitude, pulse width and frequency. The results indicate the selection criteria and the ranges of the stimulation parameters for individual-based treatment. Tests demonstrate that the personalized parameters of the VNS treatment for epilepsy and depression are based on the different response characteristics of the A and B components of the CAP. These results offer a foundation for improving the therapeutic efficacy and for developing closed-loop VNS therapies.
AUTOMOTIVE ENGINEERING
• Select
  Flank error correction of hypoid gears based on regression analyses

  TIAN Cheng, DING Weiqi, GUI Liangjin, FAN Zijie

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 141-146. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.005

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Tooth surface equations are derived based on the hypoid-format-tilt (HFT) method to calculate the machine parameter correction for the flank machining error of hypoid gears. The tooth surface discretization method is given for the flank machining error. The sensitivity coefficient for the flank error machine parameters is used to develop a mathematical model for the flank error correction method to identify its disadvantages based on a regression analysis. The linear dependence of the sensitivity coefficient vector of the machine parameters is compared with the real flank error vector to identify the variables for the regression analysis, which reduces the number of machine parameters to be adjusted. Tests of the flank error correction of a real gear show the above advantages and the effectiveness of this method. Guidelines are given for other flank error correction problems.
INFORMATION ENGINEERING
• Select
  Design and optimization of a low resource speech recognition system

  ZHANG Pengyuan, JI Zhe, HOU Wei, JIN Xin, HAN Weisheng

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 147-152. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.006

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Wearable devices and smart home systems need speech recognition engines with few resources and high rejection rates. Traditional methods cannot provide such systems. This paper presents algorithms for decoding and rejection for a low source speech recognition system. The decoding improves the rejection rate up to 64.8% by changing the filler reentry while the memory is only increased 8.5 kB compared with the baseline system. The rejection algorithm computes a background probability which is compared to similar probabilities calculated in advance online decoding. The system gives a rejection rate of 93.8% with little loss in the recognition rate. The memory and computational speed are also optimized.
• Select
  Automatic speech recognition with robot noise

  WANG Jianrong, ZHANG Ju, LU Wenhuan, WEI Jianguo, DANG Jianwu

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 153-157. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.007

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Robots inevitably produce noise when they are moving any part of their body. Such noise is caused by the various body joint motors as well as the CPU cooling fans. Moreover, these noises are easily captured by the robots' microphones because they are closer to the microphones than the target speech source. This paper presents a de-noising method using the spectral subtraction, joint noise template substraction, labeled area cepstral mean substraction and multi-condition training to estimate and suppress robot noise. Tests show that this method significantly reduces the effect of robot noise which enhances the automatic speech recognition.
• Select
  Modeling of the tongue tip in Standard Chinese using MRI

  WANG Gaowu, DANG Jianwu, KONG Jiangping

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 158-163. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.008

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  The tongue tip motion in Standard Chinese was modeled based on articulatory data from magnetic resonance imaging (MRI) images. An MRI articulatory database was developed for Standard Chinese, including 9 vowels and 75 consonant variants. Principle component analysis (PCA) of the tongue shape was then used to find articulatory factors. The results show that the tongue should be divided as the tongue tip and tongue body and modeled separately for more precise results. The tongue tip motion is modeled with two articulatory parameters for tongue tip protrude and tongue tip raise which represent the protruding/advancing and raising/retroflexing movements of the tongue tip.
• Select
  Influences of vowels on the perception of nasal codas in Mandarin for Japanese and Chinese natives

  ZHANG Jinsong, WANG Zuyan

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 164-169. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.009

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Japanese natives often have difficulties discriminating between the Chinese alveolar and velar finals. The perceptual differences between Japanese and Chinese natives are studied using different speech stimuli synthesized by concatenating speech segments including the vowel, the nasalized vowel and the nasal coda in different ways for use in a series of sound discriminating experiments. The results show that the nasalized vowels play key roles in cueing Chinese subjects to judge the nasal coda but have less effect for Japanese subjects. When the nasalized vowel portions are missing, the vowel nuclei have similar influences on the perceptions of the Chinese and Japanese. Larger acoustic differences between the vowel nuclei for alveolar/velar pairs enable both the Chinese and Japanese subjects to more accurately distinguish them. These findings can clarify the perceptual cues for Chinese nasal finals and will lead to more efficient methods to teach Chinese to Japanese students.
• Select
  Tone training for Mandarin two-syllable words based on pitch projection synthesized speech

  XIE Yanlu, ZHANG Bei, ZHANG Jinsong

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 170-175. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.010

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  This study uses the pitch projection method to synthesize teaching speech with the appropriate standard voice. The teaching speech is synthesized by turning lexicon tones in the learners' speech into standard tones, while keeping the segments and timbie unchanged. This simplifies the complex variations in the speech signal except for the tones. Then, the system is used for tone training Japanese students based on the synthesized Mandarin two-syllable words. The training results show that this synthesized speech method is superior to a standard voice method with improved perception and production, as well as generalized production. The training results for the synthesized speech method are far better than a control group without training. Most of the results are statistically significant. Tests also show the existence of a selective attention mechanism in the human brain when learning speech. Thus, this study provides an experimental and theoretical basis for speech synthesized methods to be integrated into computer-assisted Mandarin tone learning systems.
• Select
  Chest and abdomen breathing while reading literature in Mandarin

  YANG Feng, KONG Jiangping

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 176-181. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.011

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Chest and abdomen breathing characteristics were studied while reading Mandarin Chinese literature. The speech and breathing signals were recorded from four annoucers while reading 90 passages with different styles, including modern poems, ci, news, and essays. The extracted breathing parameters were the breathing reset amplitude, duration, slope, and area during the inhalation and exhalation phases. The results show that the breathing strategy during speech differs from that during natural breathing with different chest and abdomen breathing patterns for different styles of prose. Systematic breathing resets occur while reading poems and ci. Three levels of breathing resets are identified while reading news and essays at the paragraph-initial, main clause-initial and subordinate clause-initial positions. The chest and abdomen breathing are coordinated during all the reading activities, but each has its distinct functions. These results are of significance to prosody and speech production and provide a basis for articulatory modeling of breathing in speech production.
• Select
  THUYG-20: A free Uyghur speech database

  Aisikaer Rouzi, YIN Shi, ZHANG Zhiyong, WANG Dong, Askar Hamdulla, ZHENG Fang

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 182-187. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.012

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Speech data plays a fundamental role in research on speech recognition. However, there are few open speech databases available for researchers in China, especially for minor languages such as Uyghur. This paper develops a Uyghur continuous speech database which is totally open and free. The database consists of 20 h of training speech and 1 h of test speech, as well as all the resources needed to construct a full Uyghur speech recognition system, including a phone set, lexicon, and text data. A recipe used to construct the baseline system is also described with results for two test sets involving clean speech and noisy speech. This paper provides a standard database for Uyghur speech recognition.
• Select
  Name recognition in the Uyghur language based on fuzzy matching and syllable-character conversion

  Abdurahim Mahmoud, Hussein Yusuf, ZHANG Jiajun, ZONG Chengqing, Askar Hamdulla

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 188-196. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.013

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Uyghur is a very agglutinative language which belongs to the Altaic family of languages with a very complex morphology. Uyghur names have many origins, so they are difficult to analyze and recognize. Thus, there is no well-developed toolkit for name recognition in Uyghur. An investigation of a large Uyghur text shows that 83% of all the names are either Uyghur names or Chinese names. Therefore, this work focuses on these two kinds of names with specific solutions for recognizing them in Uyghur texts. A letter-based fuzzy matching method is used for the Uyghur names with a syllable-character conversion method based on a machine translation method for the Chinese names. Tests show that this method achieves a 91.84% F₁ score for the Uyghur names and 95.86% for the Chinese names.
• Select
  Emotion recognition from Uyghur sentences based on combinations of class discrimination words and a sentiment dictionary

  Abdusalam Dawut, Hussein Yusuf, Askar Hamdulla

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 197-201. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.014

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  This paper presents a recognition method for Uyghur sentence sentiments, such as anger, happiness, sadness and wonder based on combining class-discrimination words (CDW) and a sentiment dictionary. The sentiment expression characteristics in the Uyghur sentence text are identified from features extracted using a CDW feature selection method for the emotion recognition. A set of emotional words is collected manually and put into a sentiment dictionary which is combined with the CDW feature words for the emotion recognition. Tests show that the combined method is more effective than only the CDW feature based method.
• Select
  Describing and predicting affective messages for expressive speech synthesis

  GAO Yingying, ZHU Weibin

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 202-207. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.015

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  A multi-perspective emotion model is presented to provide more details about the emotions in expressive speech synthesis and to facilitate automatic predictions. The method describes the emotion development in terms of the cognitive appraisal, psychological feeling, physical response and utterance manner. The descriptive model is used to develop a text-based emotion prediction model using a deep neural network (the deep stacking network), which supports distributed representation and has a stacking structure. Tests validate the benefits of using this prediction method for the interactions among different emotional aspects and the contextual impacts, as well as the effectiveness of the deep stacking network and the multi-perspective emotion model.
• Select
  Cloud service performance bottleneck diagnosis based on the component service quality and performance

  GUO Jun, MA Anxiang, YAN Yongming, MENG Yu, ZHANG Bin

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 208-212. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.016

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Bottlenecks in component services need to be identified to ensure the performance of cloud service system-oriented service business processes. Traditional approaches for evaluating component service bottlenecks often evaluate the maximum run time delay in the component services to determine the cause of the quality-of-service deterioration. However, these approaches do not consider the importance of the component services, which influences the evaluation accuracy. A cloud service bottleneck diagnostic method is given here based on the quality of service on the various components in the analysis for comprehensive assessments of the quality of service and the component importance to identify cloud service bottlenecks in component services. Simulations show the effectiveness and accuracy of this bottleneck diagnosis method.
• Select
  Energy-efficient cluster-based privacy data aggregation for wireless sensor networks

  MAN Dapeng, WANG Chenye, YANG Wu, WANG Wei, XUAN Shichang, JIN Xiaopeng

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 213-219. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.017

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Current privacy-preserving data aggregation methods in wireless sensor networks often have large computational and communication costs. This paper presents an energy-efficient cluster-based privacy data aggregation (E-CPDA) mechanism based on the cluster-based privacy data aggregation (CPDA) method. In each round of aggregation, the cluster head chooses a node as a collaborative node for the aggregation, which reduces the computational and communication costs between the nodes in one cluster. Simulations show that E-CPDA has less communication and computational costs with good privacy-preserving performance and higher accuracy than CPDA.
• Select
  Similar trajectory query method based on massive vehicle license plate recognition data

  ZHAO Zhuofeng, LU Shuai, HAN Yanbo

  Journal of Tsinghua University(Science and Technology). 2017, 57(2): 220-224. https://doi.org/10.16511/j.cnki.qhdxxb.2017.22.018

  Abstract ( ) Download PDF ( )   Knowledge map   Save

  Vehicle license plate recognition data provides a kind of traffic monitoring data that is a large spatial-temporal stream with fixed positions. Similar trajectory queries of such data face several problems. This paper presents a similar trajectory query method based on site companions with multistage task parallelization based on the MapReduce computing model. This method gives more efficient similar trajectory queries in a distributed computing environment for massive license plate recognition data. Tests show that this method can correctly query similar trajectories more efficiently than traditional stand-alone methods based on tests with almost ten million real vehicle license plate data points.