The increasing demand of energy density and charging rate of lithium- ion batteries rise concerns about safety hazard. Critical battery failures, such as thermal runaway, can lead to explosions and pose serious safety risks for users. In this work, we developed a novel method for early fault detection in lithium-ion batteries based on the spectral analysis of the cycling test signals. We analyzed voltage and current of cycling tests in lithium-ion batteries with silicon base anodes. The spectral analysis is performed by the Fourier and wavelet transforms (continuous, cross and coherence). Results show that prior to an irregular event or critical failure, new frequencies appear superimposed on the main cycling one, signaling the onset of degradation or failure. We concluded that spectral analysis of the cycling signal enables the early detection of battery failures. This method could be implemented in battery management systems to reduce the risk of critical failures.