| 摘要: |
| 目的 声音是产品和用户之间的一种沟通媒介，为了增进设计师对产品声音的理解、合成与设计匹配，提出一种交互式可视化产品声音数据聚类分析框架。方法 首先通过神经网络将设计师感官描述式信息与声音的特征参数进行融合嵌套;其次基于高斯混合模型来描述非线性几何分布的产品声音数据;最后设计师输入个人先验知识经验参与交互聚类。结果 基于Python的Anaconda3包开发了产品声音交互式聚类的可视化分析实验工具，得到最优化产品声音聚类结果。结论 该产品声音交互聚类可视化分析工具融合了声音技术参数和人脑听觉反应机制，在聚类过程中允许用户参与交互并融入用户的先验知识，并行视图可以实时显示数据元素的流向和判别类别的稳定性。同时，可视化分析可以帮助用户横向比较各聚类结果的异同，样本的比例分布与合理性，以期寻求最优聚类结果。 |
| 关键词: 产品声音 融合嵌套 交互式聚类 可视化分析 |
| DOI：10.19554/j.cnki.1001-3563.2023.08.011 |
| 分类号:TB472 |
| 基金项目:宁波市哲社规划课题(G21-3-ZX75);浙江省基础公益研究计划项目(LGG19E050004) |
|
| Interactive Clustering Design of Product Sound |
|
ZHANG Yang, CHEN Wen-ying, PI Shan, DING Sheng-nian
|
|
(Ningbo University of Finance & Economic, Zhejiang Ningbo 315175, China)
|
| Abstract: |
| In view of that the sound is a communication medium between product and user, the work aims to propose an interactive visual analysis framework of product sound data to improve the designers' capability of understanding, synthesizing, designing and matching the product sound. Firstly, the sensory description information of designers was fused and integrated with the characteristic parameters of sound through neural network. Secondly, gaussian mixture model was used to describe the product sound data in nonlinear geometric distribution. Finally, the designers input personal prior knowledge and experience to participate in interactive clustering. Based on Anaconda3 of Python, a visual analysis experimental tool for interactive product sound clustering was developed, and the optimal product sound clustering results were obtained. The visual analysis tool for interactive clustering of product sound combines the technical parameters of sound and the auditory reaction mechanism of human brain, allowing users to participate in interaction and integrate their prior knowledge in the clustering process. The parallel view can display the flow direction of data elements and judge the stability of categories in real time. At the same time, visual analysis can help users to compare the similarities and differences of clustering results horizontally, the proportional distribution and rationality of samples, in order to seek the best clustering results. |
| Key words: product sound fuse and integrate interactive clustering visualization analysis |
