As a newly developed pair-breaking superconducting detector, microwave kinetic inductance detectors (MKIDs) are simple to integrate in the frequency domain and have already been used in astronomical detection and array imaging at the (sub)millimeter and optical wavelengths. For these applications, the dark noise level of kinetic inductance detectors is one of the key performance indicators. Here we give a detailed introduction to a noise power spectrum analysis method which can perform accurate and effcient noise spectrum analysis in a wide frequency range for kinetic inductance detectors. This method can well balance the noise spectrum resolution and variance performance, by taking the noise data at the resonance frequency with two sampling rates and setting appropriate frequency resolutions for different frequency bands. We use this method to characterize and compare the noise of aluminum (Al) kinetic inductance detectors made from two different micro-fabrication processes. We deposite a 25 nm thick aluminum film on high-resistivity silicon substrate for one device, while place silicon nitride (SiN<sub><i>x</i></sub>) films on both the top and bottom of the aluminum film for another device. It is found that the frequency noise of the device with double silicon nitride films is approximately 25% to 50% of the bare aluminum device. Based on this double silicon nitride film fabrication technique, we further fabricate a few groups of lumped-element aluminum kinetic inductance detectors with various inductor and interdigitated capacitor (IDC) designs. We investigate the noise properties of these devices at different microwave driven powers and bath temperatures, and the experimental results show typical two-level system (TLS) noise behaviors. Our work provides a standard method to characterize the noise power spectrum of kinetic inductor detectors, and also paves the way to develop low-noise aluminum kinetic inductor detectors for terahertz imaging, photon-counting and energyresolving applications.