Lithium-ion battery plays an increasingly important role in automotive and mobile power systems due to its relatively high energy density, low self-discharge rate, and long cycle lifetime. The analysis of electrochemical impedance spectroscopy (EIS) data is essential for both characterization and fault diagnosis applications in managing lithium-ion battery cells. While several studies on state of charge (SOC) estimation using impedance data have been reported, it is still largely unclear regarding the feasibility of the state of health (SOH) estimation using the EIS data and further how capacity fade would affect the estimation. This study aims to analyze the impedance data using an empirical approach to estimate the SOC and SOH. Several battery cycling profiles are designed to ensure that the EIS measurements are taken under different operating conditions. For the SOC estimation, impedance measurements are recorded with a pace of 10% capacity drops. Meanwhile, for the SOH estimation, battery cells are cycled until the end of life with accelerated aging test. The result of the correlation analysis shows that the middle frequency section of the Nyquist plot has a strong correlation with the SOH. Additionally, correlation analysis is used to select input data from the Nyquist plot when the battery is at different SOC level. The relationships between the impedance value and both SOC and SOH are exploited by feeding the selected data through a recurrent neural network (RNN). Batteries under different operating conditions were tested using the developed technique to demonstrate the efficacy in accurate estimations of both SOC and SOH.