Wood material recognition for industrial applications
Published in Systems Science & Control Engineering, 2018
Haibin Fu, Huaping Liu, Xiaoyan Deng, Fuchun Sun
At present, there are many studies on sound recognition. Feature extraction plays an important role in sound recognition. According to the survey, the extraction methods are mainly the zero crossing rate (Ghosal & Chakraborty, 2009), the time and frequency domain analysis (Pai, Deng, & Sundaresan, 2015), the Linear Prediction cepstral coefficients (LPCC) (Yusnita, Paulraj, & Yaacob, 2011), wavelet transform (Feizifar, Haghifam, & Soleymani, 2012), the Mel Frequency Cepstral Coefficients (Ahmad, Thosar, & Nirmal, 2015) and others. The choice of classification algorithms also has a crucial role for sound recognition, such as artificial neural networks (ANN) (Gao, 2012), extreme learning machine (ELM) (Huang, Zhu, & Siew, 2004), Gaussian Mixture Model (GMM) (Mohammadi & Saeidi, 2008), etc. Based on the fast learning speed and computational efficiency, ELM is more flexible and computationally attractive than traditional learning methods (Wei, Liu, Yan, & Sun, 2016).