基于YAMNet迁移学习的智能手术语音识别研究
Research on Intelligent Surgical Speech Recognition Based on YAMNet Transfer Learning
孙文业 1徐伟 2汪春亮1
作者信息
- 1. 苏州大学附属第二医院,江苏 苏州 215004
- 2. 常熟理工学院,江苏 苏州 215500
- 折叠
摘要
随着人工智能技术的不断进步,语音识别技术作为人工智能技术领域的一项关键技术,可以帮助医生和护士更高效的沟通和操作.文章提出了一种基于YAMNet模型迁移学习网络的智能手术语音识别方法,通过迁移学习技术将YAMNet模型应用于手术语音识别任务中,实现对手术器械清点过程中语音的自动识别.首先收集了手术清点常用器械语音数据,其次利用训练集对迁移学习后的网络模型进行训练,最后通过验证集对该网络模型进行验证.实验结果表明,所提方法在智能手术语音识别任务中取得了显著的性能,识别准确率达到 97%,为智能手术系统的发展提供了新的思路和方法.
Abstract
With the continuous advancements in Artificial Intelligence technology,as a key technology in Artificial Intelligence technology field,speech recognition technology can help doctors and nurses communicate and operate more efficiently.This paper proposes an intelligent surgical speech recognition method based on the YAMNet model and Transfer Learning network.By Transfer Learning technology,it applies the YAMNet model to surgical speech recognition tasks,realizing automatic speech recognition during the counting of surgical instruments.Firstly,speech data of commonly used surgical instruments during the counting process is collected.Secondly,the network model after Transfer Learning,is trained using the training set.Finally,the network model is validated by the validation set.Experimental results show that the proposed method achieves significant performance in intelligent surgical speech recognition tasks,with a recognition accuracy rate of 97%,providing new ideas and methods for the development of intelligent surgical systems.
关键词
语音识别/YAMNet/迁移学习/手术清点Key words
speech recognition/YAMNet/Transfer Learning/counting of surgical instruments引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据