Data on the duration of traffic accidents have been collected by using a manual recoding method based on the statement of freeway patrol officers or eyewitness. Accordingly, the data may include cost and safety problems due to the difficulty of observing an accident from its occurrence to its clearance. To reduce these problems, this study proposes an analytical method to estimate accident duration based on the archived speed profile. The proposed method is demonstrated by a case study using data on approximately 6,200 accidents that occurred on freeways in Orange County, California, in 2001. In addition, a statistical analysis based on the Accelerated Failure Time (AFT) metric survival models was performed to identify the relationship between estimated accident duration and its causal factors. The analysis results were consistent with the theoretical expectation and prior studies. Ultimately, the proposed method will contribute to reducing observation burdens when collecting data on accident duration.