光谱图
噪音(视频)
谐波
时频分析
声学
计算机科学
噪声测量
源分离
语音识别
人工智能
工程类
降噪
计算机视觉
图像(数学)
物理
滤波器(信号处理)
作者
Ryugo Uchiyama,Nari Tanabe
标识
DOI:10.1109/apsipaasc58517.2023.10317304
摘要
This paper proposes a method to suppress noise in real-time using harmonic/percussive separation with morphological operations for hammering tests. The feature of the proposed method is that it focuses on the structure of the spectra of the hammering sound and noise obtained by analyzing the spectrogram of the signal observed in the hammering test and performs morphological operations using structuring elements suitable for each shape. This enables more appropriate extraction of the hammering sound than conventional methods. Although only the spectra of past time frames are used to achieve real-time noise suppression, the noise suppression performance is better than that of previous methods. The effectiveness of the proposed method is demonstrated by subjective and objective evaluations.
科研通智能强力驱动
Strongly Powered by AbleSci AI