This study aims to investigate the effect of length features of Turkish musical instruments on the quality of their recognition and classification, and to explore the ways in which using exclusively the general features of voice recognition or using exclusively the short duration or long and medium length features can provide acceptable results in the field of  Turkish musical instruments. For this purpose, after collecting musical datasets in four instrumental categories, including bow, plectrum, percussion, and woodwind, three experiments with similar conditions by using the neural network on this database are conducted to assess the effectiveness of the use of general features as well as the long, short, and medium lengths features. The result indicates that the use of the general properties and characteristics of sound, which have been used in several previous authoritative types of research, leads to the categorization with the accuracy of 59%, and the use of short-term features and long and medium length features lead to the categorization with the accuracy of 63% and 81% respectively. However, the selection of features from all four categories will lead to the categorization of 98% accuracy.