We analyze a dense cluster of solar radio spikes registered at ~ 4.5 -- 6 GHz by the Purple Mountain Observatory spectrometer (Nanjing, China) operating in the 4.5 -- 7.5 GHz range with the 5 ms temporal resolution. To handle with the data from the spectrometer we developed a new technique utilizing a nonlinear multi-Gaussian spectral fit based on chi-squared criteria to extract individual spikes from the originally recorded spectra. Applying this method to the experimental raw data we eventually identified about 3000 spikes for this event, which allows for a detailed statistical analysis. Various statistical characteristics of the spikes have been evaluated, including intensity distributions, spectral bandwidth distributions, and distribution of the spike mean frequencies. The most striking finding of this analysis is distributions of the spike bandwidth, which are remarkably asymmetric. To reveal the underlaying microphysics we explore the local trap model with the renormalized theory of spectral profile of the electron cyclotron maser (ECM) emission peak in a source with random magnetic irregularities. The distribution of the solar spikes relative bandwidth calculated within the local trap model represents an excellent fit to the experimental data. Accordingly, the developed technique may offer a new tool of studying very low levels of the magnetic turbulence in the spike sources, when the ECM mechanism of the spike cluster is confirmed.