Determination of peak locations and amplitudes for automated interpretation of spectra

Exact determination of the position and magnitude of peaks in a spectrum is essential to interpreting a spectrum, especially for automating the process. Algorithms such as FFT provide values at regular intervals, whose spacing is determined by the length of the acquired signal. If the time-domain signal length does not contain a whole number of periods, the peak position falls between two spectral values. Neither the exact position nor the amplitude are apparent from the spectrum. A method is presented here for computing a correction to both the location and the amplitude of the local maximum in the spectrogram to significantly improve the estimated location and frequency of the peak location. The computational complexity is constant time per peak, thus it can be employed in low computational power and embedded systems. Corrections for signal length and zero-padding are also included. Examples of the application of this method are presented and its accuracy vs. other methods is discussed.