An efficient algorithm for Gender Detection using voice samples

Acoustic signal, speech, having a property for detecting the gender of a speaker. This is well known as Gender Detection (GD). In this paper, we propose pitch based gender detection algorithm. Pitch is the fundamental frequency of speech signal. Gender Detection using pitch can be performed in time domain, frequency domain, or in both. In this current paper, we propose an efficient time domain based gender detection (GD) algorithm, where, we use Autocorrelation (ACF) as pitch detection technique, and K-Means as a classifier. In our algorithm, we introduce efficient frame selection techniques to determine the pitch from speech signal. This technique not only improves the performance of the pitch detection algorithms, but also it improves the classifier efficiency. Our proposed algorithm was verified over a forty speaker based speech database, and the percentage of correctness for the gender classification was measured. The performance was improved after our algorithm was implemented.