Transformation of Voice Signals to Spatial Domain for Code Optimization in Digital Image Processing

The paper is intended to transform the voice-signal from the frequency domain into a spatial domain in form of grayscale image and applied the image processing techniques. To satisfy our hypothesis, two models of signal processing were carried out in this research: Speaker Recognition and Signal Segmentation. For applying the image processing techniques on the voice-signal, two methodologies were proposed to convert the signal into grayscale-image: signal-range based and fuzzy-based. The signal-range based is to convert the signal range from (−1 ↔ 1) into (0 ↔ 256). The second method of conversion, Fuzzy Gaussian Membership Function is applied to convert the signal range into (0 ↔ 1), then multiply them by 255 to be in the range of grayscale image. In the Speaker Recognition, the LBP is used as pre-processing for filtering the intensity of the signal image. The HOG is used to extract the features of signal-image. So, the total length of features-vector is 324. The classification learner tool in MATLAB was used for classifying the feature-vectors and the results were found to be satisfactory. The automatic word segmentation was proposed based on thresholding and morphology operators. The segmentation accuracy is 93.67% in the Marathi-language. The highest recognition rate in speaker identification system is 96.9%.