Spatial Processing for Improved Quality Recognition of Optically Recorded Voice Signals and Illumination Varied Scenery
Yevgeny Beiderman, Yaniv Azani, Yoni Cohen, Chen Nisankoren, Mina Teicher, Ehud Rivlin, Vicente Mico, Javier Garcia and Zeev Zalevsky
Affiliation: School of Engineering, Bar Ilan University, Ramat Gan, 52900, Israel.
In this paper we present two image processing techniques allowing improving the recognition quality of remote optically recorded voice signals (the first technique) and of images taken under varied illumination conditions (the second technique). Regarding the first approach dealing with the processing of voice signals that were recorded by a patented optical technology, we present a concept that improves the quality of the recording and the classification of the characteristics of the recording while the proposed signal processing operations are applied over the spectrogram of the optically recorded signals. Regarding the second approach dealing with the illumination varied conditions, we present a new technique based upon combination between spectral (of colors) manipulation called the HSV and spatial transformation called the K-factor that is applied over the HSV components. Such manipulation allows composing image which is both insensitive to illumination and contains the significant spatial details of the originally imaged object.
Keywords: Optical microphone, spectrogram, classification, pattern recognition, illumination invariance, Spatial Processing, polarization, Voice Signals, invariant imaging, optical microphone system, Gaussian image, V image, optical recording technology, Computer Vision, Harris corner detection algorithm, gray level image
Rights & PermissionsPrintExport