Miniature spectrometers with desirable properties such as being small in size, light weight and non-fragile provide
solutions to a variety of promising application. In this work, through the developed algorithms, a low-cost chip-scale
spectrometer is demonstrated in particular for LED spectrum sensing, as the market of LED lights has been booming
and low-cost solution for testing and monitoring LEDs spectral characteristics become essential. The developed
algorithms are in two forms, namely algebraic approach and training approach. For the algebraic approach,
non-negative least square method is reported useful for spectrum reconstruction of narrowband LED spectra. For the
training approach, FWHM measurement and peak wavelength measurement, the two major parameters for specifying
monotonic LED spectra, are reported with accuracy within 1nm.