Hardware strategies for end-point detection

End-point detection of a speech signal is an important process in isolated-word recognition. Several approaches were suggested to isolate speech signal from the back-ground noise. One such approach is using energy and zero crossing rate. In the paper, a variation of this algorithm is considered for hardware implementations for both on-the-fly (real time applications) and off-the-fly (non-real time applications) methods. This algorithm is chosen because of its simplicity and hardware suitability for implementation. The reason for hardware implementation is to achieve better speed performance over software approaches. The speed of energy magnitude computations alone, in off-the-fly approach, is O(k); where k is the frame, of arbitrary length, count. The approach in on-the-fly method is the most suitable for end-point detection implementation for real time applications.