CASIA OpenIR  > 毕业生  > 博士学位论文
手语信息处理系统及数据融合方法研究
其他题名Research on Sign Language Information Processing System and Data Fusion Methods
马小军
学位类型工学博士
导师原魁
2001-02-01
学位授予单位中国科学院研究生院
学位授予地点中国科学院自动化研究所
学位专业控制理论与控制工程
关键词虚拟现实 多传感器融合 数据手套 弯曲传感器 模糊融合 模糊测度 D-s证据理论 Virtual Reality Sensor Fusion Dataglove Bend Sensor Fuzzy Fusion Fuzzy Measure D-s Evidential Theory
摘要手语处理技术是一门新兴的应用技术,数据手套是手语信息系统的重要设 备之一。本文主要围绕数据手套的研制及数据手套的信息融合展开研究。在国 家“八六三”高技术项目“高性能、低成本数据手套的研制”的资助下,调研 和分析了目前国内外数据手套的类型及其传感器的特点,研制了新型传感器及 数据手套,并且对数据手套不同层次的信息进行融合方法的研究。本论文研究 宗旨是使数据手套尽量低成本、高性能,融合算法尽量运算速度快,从而保证 数据手套的实时性。因此论文包括以下几方面的工作: 首先研制出新型弯曲传感器,并对其进行理论分析和实验研究,得到新型 弯曲传感器的角度特性。传感器的制作材料和结构不但使它具有较低的成本和 高可靠性,而且还使其具有柔性好、重量轻等明显特点,这可以保证测量部位 的准确性和测量的充分性。该传感器主要用于测量物体表面弯曲角度或两个物 体之间的夹角,若将该传感器与计算机采样系统配合使用,可以很方便地实现 给定对象的动态测量。新型传感器特别适用于对精度要求不太高但需要对角度 进行连续和快速测量的场合。 其次,在新型弯曲传感器的基础上,研制了一种新型数据手套CAS_Glove。 该数据手套的主要作用是使操作者能以一种自然的方式将手的位姿信息传递给 计算机系统,可以作为手信息输入设备应用于虚拟现实系统或手语识别系统中。 这种数据手套的一个重要特点是,测量的手部信息实时性好。我们还研制了与 该数据手套配套使用的信号转换电路以及A/D采样电路,并且通过对该数据手 套上传感器的A/D采样数据分析可得出,采样转换值经处理可以完成对角度的 不精确测量,反映出手的位姿。 第三,数据手套作为手语信息输入系统的一个子系统,本文对其底层信息 进行模糊融合研究。1、根据实验数据特点确定了数据手套底层信息融合时的相 容性关系、模糊测度和IOWA算子。采用模糊建模的思路动态确定底层融合函 数中的融合参数Q。2、对标定信息进行融合。3、针对数据手套上单传感器的 角度特点,使用一种改进的期望输出模糊隶属度法将传感器的保守角度和风险 角度进行融合。实验数据表明了这些融合策略对于数据手套底层融合的合理性。 第四,在分析人手模型和30个手指语拼音字母基础上,提出采用弯曲和外 展传感器分别融合的策略,并且以弯曲传感器为主,外展为辅。根据融合中的 不确定性,提出一种将弯曲传感器角度分类融合的结合型算子。对比实验表明该算子在进行数据手套特
其他摘要The technology of sign language processing is a practical technology around the world. Dataglove is an important device of sign language information system. The thesis mainly deals with developing dataglove and data fusion of this dataglove. Under the support of "863" national high-tech project "Researching and Developing a Kind of High Performance and Low Cost Dataglove", styles of datagloves and characteristics of sensors are investigated and analyzed. In this thesis, new kinds of bend sensor and dataglove are developed and the methods of their data fusion are deeply researched. Intention of the research is to guarantee low cost and high performance of the dataglove and real-time of data fusion methods. The main work and contribution of this thesis are as follows: Firstly, a new kind of bend sensor is developed. Angle characteristics of the sensor are gotten by experiment and analyzing. The manufacturing materials and structure of the sensor make it own characteristics of not only high reliability but also flexibility and lightweight. Its characteristics ensure the veracity and sufficiency of measuring. It can be used to detect the bend angle of an object or inclination between two objects. When the sensor is used with a proper converting circuit, it can implement dynamic measuring the bend angle of an object conveniently. It is especially suited for the kind of task in which the angle needs not to be measured with high precision but needs to be measured continuously and in real-time. Secondly, based on the new sensor, a new kind of dataglove named CAS Glove is developed. Its basic function is to measure the movement of user's hand in real- time and transfer the information to computer system. It can be used as input device of sign language in virtual reality system or sign language recognition system. One of important features of this datagove is the real time of measuring. The signal converting circuit and A/D sampling circuit of the dataglove are also developed. Experiments indicate this dataglove can complete the imprecision measuring of fingers' angles. Thirdly, for the datagove is one sub-system of sign language information system, its bottom layer information fusion is studied. 1. Considering the experiment data, we establish compatibility relation, a kind of fuzzy measure and IOWA operator of the dataglove's bottom information fusion. Parameter a of bottom fusion function can be gotten by real-time data based on fuzzy techniques. 2. Label information is fused. 3. To the specialties of the sensor in the dataglove, conservative angle and risk angle are fused by improved method of EOMF ( expected output membership function ). Experiment data show these fusion techniques are suitable for bottom data fusion of the dataglove. Fourthly, bend sensors fusion and abduction sensors fusion are executed respectively according to the specialties of hand model and 30 Chinese basic letters. Based on the uncertaintie
馆藏号XWLW650
其他标识符650
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/5715
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
马小军. 手语信息处理系统及数据融合方法研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2001.
条目包含的文件
条目无相关文件。
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[马小军]的文章
百度学术
百度学术中相似的文章
[马小军]的文章
必应学术
必应学术中相似的文章
[马小军]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。