This paper proposes novel robust speech F₀ estimation using Summation Residual Harmonics (SRH) based on TV-CAR (Time-Varying Complex AR) analysis. The SRH-based F₀ estimation was proposed by A. Alwan, in which the criterion is calculated from LP residual signals. The criterion is summation of residual spectrum value for harmonics. In this paper, we propose SRH-based F₀ estimation based on the TV-CAR analysis, in which the criterion is calculated from the complex AR residual. Since complex AR residual provides higher resolution of spectrum, it can be considered that the criterion is effective for F₀ estimation. The experimental results demonstrate that the proposed method performs better than conventional methods; weighted auto-correlation and YIN.