Loading...

Pitch Detection Using Deep Learning

Khademhosseini, Mohammad | 2019

449 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 52629 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Marvasti, Farrokh; Ghaemmaghami, Shahrokh
  7. Abstract:
  8. Pitch frequency is one of the most important attributes of speech, which has been found to be quite challenging in noisy conditions. In this paper, we propose a pitch detection method based on separation of the low pitch from high pitch signals, depending on the pitch frequency below or over 200Hz, respectively, using a deep convolutional neural network. The pitch frequency is initially estimated, employing a conventional pitch detection method. From this initial estimation and using a deep convolutional neural network which determines the signals type (high-pitch or low-pitch), the pitch candidates are derived. To choose the true pitch values, we use three features in addition to soft decision of the deep network, i.e., we scale the values of the features of low-pitch and high-pitch candidates with softmax probabilities. The first feature is based on peak positions of the signal spectrum in frequency domain and the second one computed from a variant of comb filtering. The third feature varies with the method used for calculating initial estimation of pitch frequency. We add these features to the pitch smoothness cost function that is used to discriminate between the pitch candidates. The simulation results on CSTR and KEELE databases, in noisy environments show that our method outperforms the state-of-the-art methods under different SNR conditions
  9. Keywords:
  10. Deep Learning ; Voice Processing ; Convolutional Neural Network ; Pitch Detection

 Digital Object List

 Bookmark

No TOC