Li-ion Batteries are one of the most critical components of the next generation Hybrid Electric Vehicles (HEV) as degradation or failure of the Li-ion battery could lead to reduced performance, operational impairment and even catastrophic safety issues. An effective diagnostics and prognostics system for Li-ion battery health monitoring would greatly improve the reliability of such systems and thus secure general public acceptance. This paper presents a similarity-based health assessment method for Li-ion battery. Instead of physically diagnosing the health of the Li-ion battery, the proposed method defines the healthy operations (charging and discharging) as the baseline and the deviation from this baseline is treated as the degradation. Specifically, novel features are extracted from the voltage, current and temperature measurements firstly. Then Principal Component Analysis (PCA) is applied to minimize the dimensionality of the multivariate feature space. Based on the principal components projected, the Gaussian Mixture Model (GMM) is built using Expectation-Maximization (EM) method. The performance degradation, as represented by Confidence Value (CV) ranging from 0 to 1, is assessed by the comparison between training mixture distribution and testing mixture distribution. The main advantages of the proposed method include: 1) the health indicator (CV) is the combinational result of all relevant features, and therefore more reliable to conclude to what extent the system has degraded; 2) little expert knowledge about Li-ion batteries is needed to implement the method; 3) the features are extracted from common exterior measurements, such as voltage, current and temperature, which are easy to acquire.