Vibration limit is one of the most essential contents in vibration serviceability research. Former studies showed that many factors, such as biological and environmental factors, significantly affected vibration limits deeply. As a reason of defects in traditional research, such as small scale data and unreal test environment, quantitative relationships between vibration limits and these factors stayed unknown. Based on data collected by crowd sensing in real environment, crest factor of vibration/ BMI/ human age/ floor of building were found key factors by using maximal information coefficient (MIC) in coefficient analysis. Functional relationship and 95% confidence intervals between vibration limits and key factors were proposed, respectively. Lilliefors test and normal probability plot show that residuals between fitted values of limits and measured ones follow a normal distribution. A novel approach of estimating vibration serviceability based on probability is proposed when key factors and vibration magnitude are known.