A new algorithm for pattern recognition of voices

An algorithm is developed to distinguish between the speech of male and female. A simple hardware device is devised to implement the algorithm. Various techniques, such as fast Fourier transform (FFT), threshold crossings (TCs), threshold crossings of the derivative of the signal, and sine-wave crossings are used to distinguish between the voices. These techniques are analyzed and compared. The results show that the sine-wave crossings technique is simple to implement, does not need normalization or differentiation of the signal to remove its DC component, and can be used for real time (online) analysis.<
>