Date ( Phoenix dactylifera ) is an ancient fruit containing great source of minerals and nutritive materials. Date palm is also one of the most important trees in desert. Iran has the largest fields of planting and ranks second grade in terms of producing date. This product plays an important role in foreign exchange income. Grading of date fruit is currently done manually which causes mechanical damage to the fruit during the operation. Therefore, developing an automatic grading system in order to increase the quality of the product and reduce the labor cost is of great importance. In this research an electro-mechanical system was developed for online measuring firmness of date fruit based on nondestructive impact force. The samples of Mazzafati variety were prepared form Jiroft (Kerman provience) at four different maturity stages (Kimri, Khalal, Rutab andTamr). The nondestructive tests were performed on two sides of the samples at three different speeds (1, 1.5 and 2 m s -1 ). Since the firmness of date fruit is related to the maturity stage, this property was considered as the grading index. Each fruit was divided into two equal parts and the core was removed. Destructive tests were also performed on both sides separately based on Magness-Taylor (MT) test using a universal tension compression device with a probe of 4 mm diameter. The maximum force of penetration was considered as the firmness index. Three different analysis methods were used to obtain a functional relationship between the firmness as the dependent variable and the recorded signal as the independent variable. In the first method, three characteristics of the signal’s first half-wave were defined including peak force (F p ), its duration (P d ), and the area under its diagram (I g ). The non-linear multivariate regression approach was performed on MT firmness against these properties. The results showed no accepatable results. The partial least squares (PLS) regression was then applied to the total impact signal data which resulted in a better firmness prediction. The calibration coefficients were also revealed that only up to a specific duration of the total signal data (200 ms) were effecitive on firmness prediction. In other words, the data after this period could be considered as noise. Consequently, a new linear multivariate regression approach considering the above mentioned characteristics (F p , P d and I g ) for the first 20 half-waves of the signal was applied for predicting firmness. The results showed an appropriate improvement in prediction confirming the applied assumption. According to the similar results for both sides separately and considering all data together, it can be concluded that firmness prediction was not dependent to the oreientation of fruit during the impact, hence the developed approach can be used for practical grading purposes. Overall, regarding enginireeing considerations related to grading system and the fact that the lower speed causes less damage to the fruit the speed of 1.45 m s -1 is recommended for developing the commercial system. Key words: Date palm, Firmness, Impact, Nondestructive test, Partial least squares, Rheological properties