Research of decision function sensitivity during frequency band detection under noise influence

Voice signal analysis and processing techniques are main directions in the field of voice technologies. Here the frequency analysis is widely used for voice signal processing. Various voice sounds differ by distribution of energy along a frequency axis. At the same time, there is a good reason to consider only frequency bands that possess the main energy of signal division that is so-called information frequency band. We propose to apply method for information frequency band detection based on comparison with adaptively defined threshold. It is clear that additive noise will change energy distribution over frequency domain. This work presents results of researches of information frequency band wrong detection probability value dependence on signal-to-noise ratio. In addition, we present results of comparison between proposed method and method based on energy part setting. Those obtained research results demonstrate the method based on using of adaptive threshold is resistant to additive noise.