Improved Windowing Method for Achieving Consistent Spectral Analysis
Original Publication Date: 1985-Sep-01
Included in the Prior Art Database: 2005-Feb-19
In various applications --such as speech recognition, compression, or analysis-- a digital signal may be analyzed to provide spectral energy information. In such analysis, a "window" is typically defined which extends over a frame of time (for example, 20 milliseconds) and which is displaced periodically a predefined time (for example, 10 milliseconds). For each time frame, a Fast Fourier Transform (FFT) is performed to indicate how the energy is distributed in predefined frequency bands. Fig. 1 is an illustration of the conventional method of displacing windows at equal successive intervals. It is noted that some of the windows occur where there is little energy whereas other windows extend over periods of high energy. When FFT is performed on windows having such different energy patterns, inconsistent information results.