Human fall causes injuries and may even lead to death in the case of older age. Due to increasing elderly population every year to the total population and the health problems and risks caused by fall especially among the age group of 60 and above, detecting fall at the earliest is essential in order to avoid human loss. Basically, fall detection is considered as a classification problem which requires developing a classifier model that recognizes and classifies normal human activities and abnormal activity like fall. Most of the existing fall detection methods are based on classifiers constructed using traditional methods such as decision trees, Bayesian Networks, Support Vector Machine etc. These classifiers may miss to cover certain hidden and interesting patterns in the data and thus suffer high false positives rates. This paper presents a novel algorithm called Frequent Bit Pattern based Associative Classification (FBPAC) that maps the tri-axial accelerometer data streams to bit patterns and mines the frequent bit pattern occurring for normal activities like sitting/standing, lying and walking within a time-sensitive sliding window. Unlike normal activities, fall have significant peak acceleration and it is detected by setting most significant bit of bit pattern and thus clearly distinguishes fall from lying activity, thereby reducing false positive rates. Empirical studies are conducted by collecting real time tri-axial accelerometer data from a wearable and unobtrusive sensing device. Experimental results show that within a time-sensitive sliding window of 10seconds, the proposed algorithm achieves up to 92% overall accuracy.