Guaranteed estimation of speech fundamental frequency with bounded complexity algorithm
Aleksandr Melnikov, Andrey Barabanov
he first stage of the Pitch estimation for a short time speech signal analysis consists of an exhaustive search of all integer Pitch periods in the harmonic model. For each Pitch period the cost function is calculated and then optimized. The general complexity is proportional to N2 where N is the length of the window in samples.
It is shown in this paper that the cost function can be approximated with the geometric rate of accuracy while the complexity of the algorithm remains around N log2 N. The trade-off between accuracy and complexity is discussed.
CYBERNETICS AND PHYSICS, Vol. 5, No. 1, 2016, 21–35.