• Sonuç bulunamadı

LIST OF FIGURES

N/A
N/A
Protected

Academic year: 2021

Share "LIST OF FIGURES"

Copied!
2
0
0

Yükleniyor.... (view fulltext now)

Tam metin

(1)

LIST OF FIGURES

Figure 2.1: Speech Recognition technology in modern cars . . . 4

Figure 2.2: Two robots controlled by speech . . . 4

Figure 2.3: Speech password device . . . 5

Figure 2.4: Speech Recognition feature in Windows 8 . . . 5

Figure 2.5: Speech signal for the word “one” . . . 7

Figure 2.6: Human speech production apparatus . . . . . . 8

Figure 3.1: General structure of Speech Recognition system . . . 12

Figure 3.2: Sampling process . . . .. . . 13

Figure 3.3: Framing process . . . 15

Figure 3.4: Block diagram of LPC estimation . . . 17

Figure 3.5: Block diagram of LPC analysis . . . 18

Figure3.6: Block diagram of LPC synthesis . . . 19

Figure3.7: Mel scale filter bank . . . 20

Figure3.8: Mel scale plot . . . . . . 20

Figure3.9: Block diagram of MFCC . . . 21

Figure3.10: Spectrogram of wideband speech signal . . . . . . 23

Figure3.11: Spectrogram of narrowband speech signal . . . . . . 24

Figure3.12: Exchange the axis of the frame . . . . . . 24

Figure3.13: Steps to get Spectrogram . . . 25

Figure 4.1: General structure of neural network . . . 27

Figure 4.2: Neural network topologies . . . 28

Figure 4.3: Mathematical neuron . . . . . . 30

Figure 4.4: Linear function . . . .. . . 31

Figure 4.5: Binary function . . . 31

Figure 4.6: Sigmoid function . . . 32

Figure 4.7: Fully connected Feed-Forward neural network. . . .. . . 33

Figure 5.1: General structure of the program . . . 36

Figure 5.2: Flowchart of LPC method . . . 38

Figure 5.3: Flowchart of MFCC method . . . 39

Figure 5.4: Flowchart of Spectrogram method . . . 40

Figure 5.5: Operation of end points detection (a) Source signal, (b) End points detected signal. 41

x

(2)

Figure 5.6: A one frame of the word “one” . . . 41

Figure 5.7: Frame signal after applying hamming window. . . 42

Figure 5.8: LPC coefficients for a one frame . . . 47

Figure 5.9: LPC coefficients for a word “one” . . . 47

Figure 5.10: Spectrum of a one windowed frame . . . 48

Figure 5.11: Mel filters bank . . . 49

Figure 5.12: MFCC coefficients for a one frame . . . 49

Figure 5.13: MFCC coefficients for the word “one” . . . 50

Figure 5.14: Obtained signal after rotating spectrum of the windowed frame . . . 51

Figure 5.15: Spectrogram of the word “one” . . . 51

Figure5.16: Speech Recognition system . . . .. . . 53

Figure 5.17: Hamming window for a one frame of the word “one” . . . 59

Figure 5.18: Hamming window of the word “one” .. . . 59

Figure 5.19: LPC coefficients for “one” . . . 60

Figure5.20: How features enter to the neural network and how output is gotten . . . 62

Figure5.21: Neural network applied window . . . . . . 64

Figure5.22: First display of displaying interface window. . . .. . . 65

Figure 5.23: Select a type of recognizing window . . . .. . . 66

Figure5.25: Selecting a not trained word window . . . 67

Figure5.25: Decision window for a trained word . . . .. . . 67

Figure 5.26: Plotting of the matched trained word . . . .. . . 68

Figure5.27: Selecting a not trained word window . . . 69

Figure 5.28: Displaying new buttons in main window . . . . . . 69

Figure 5.29: Speech signal after adding noise . . . 71

xi

Referanslar

Benzer Belgeler

SEM micrographs of sol-Au nanoparticles (left) and sol-LiF nanoparticles (right) after a single deposition on ITO coated glass. The scale bar represents 200 nm. Ow-Yang,

Keywords: angular motion estimation, Euler angles, angular rates, IMU, sensor fusion, Kalman filter, ballbot, stabilization, acceleration control, balancing control Reliable

While the search server stores the secure index generated by the data owner as explained in Section 6.3.1, the file server stores the actual encrypted data elements and knows

Until now, we reviewed that scale free networks have almost power-law distributed degree, shortest-path betweenness and random walk betweenness distributions, and

Figure 3.6 Overall Raman scattering spectra of pristine and 16 hour-ozone treated samples including OD-MWNTs, OW-MWNTs, SOD-MWNTs. 43 Figure 3.9 Change in surface composition of

Table 5.1: The time consumed by a hologram reconstruction using angular spectrum method and the time consumed by a sharpness estimation using normalized-variance and

Author contributed to the research in various aspects such as overall mechanical design of the microfactory system, de- sign and precise control of three particular functional

In Figure 5.3 each technique’s performance is visualized and compared against actual values. We have normalized the answers as judged distance / real distance then removed outliers