权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

DEVELOPMENT OF SCREENING SYSTEM FOR LARYNGEAL CANCER USING SPEECH INFORMATION THROUGH TELEPHONE LINES

利用电话线路语音信息开发喉癌筛查系统

基本信息

批准号：
07555126
负责人：
KASUYA Hideki
金额：
$ 2.5万
依托单位：
UTSUNOMIYA UNIVERSITY
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (B)
财政年份：
1995
资助国家：
日本
起止时间：
1995 至 1996
项目状态：
已结题

来源：
https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-07555126/
关键词：
laryngeal cancer speech analysis telephone voice hoarseness

项目摘要

Database of pathological voice has been created, which consists of 345 vowels sustained by patients with laryngeal cancer, vocal fold polyp, recurrent laryngeal nerve paralysis, vocal fold nodule, sulcus vocalis and other typical laryngeal disearses as well as by normal subjects. All the vowel signals have been recorded through the public telephone lines. The database includes information on laryngeal disease, age, gender, psychological voice quality judgments (GRBAS) of the patient together with both direct and telephone voice signals. The acoustic parameter values of jitter, shimmer, laryngeal noise, spectral envelope, and average fundamental period have been compared between the vowel signals before and after passing the telephone line. Statistical test has shown that no significant difference exists between them in terms of the screening. An artificial neural network with three layrs have been trained to classify an input vowel utterance into three categories, i.e.severely hoarse (laryngeal cancer doubted), hoarse (recommended to have check up by an ENT doctor) and no hoarseness. A screening system for laryngeal cancer has been installed at National Cancer Center Hospital East and has been tested for about 450 subjects who are mostly residents around Kashiwa City area where the hospital is located. Performance evaluation of the system is now underway.

建立了包括喉癌、声带息肉、喉返神经麻痹、声带小结、声带沟等典型喉癌患者和正常人345个元音的病理性嗓音数据库。所有的元音信号都已通过公共电话线记录下来。该数据库包括患者的喉部疾病、年龄、性别、心理语音质量判断(GRBA)以及直接语音和电话语音信号的信息。比较了元音信号通过电话线前后的抖动、抖动、喉部噪声、频谱包络和平均基频周期等声学参数值。统计检验表明，两者在筛查方面没有显着差异。训练了一个三层人工神经网络，将输入的元音语音分为三类，即严重沙哑(怀疑喉癌)、沙哑(建议耳鼻喉科医生检查)和非沙哑。喉癌筛查系统已经安装在国家癌症中心医院东部，并已在大约450名受试者中进行了测试，这些受试者大多是医院所在的柏川市附近的居民。目前正在对该系统进行绩效评估。