Non uniform TSM is performed applying various values of scaling variables for different speech units i. e. vowels, consonants and telephone transitions. Scaling factors are selected inside a way that preserves the purely natural prosody, i. e. vowels are stretched with greater components than for consonant, even though cellphone transitions continue to be intact. Depending on the input speech price, the signal is modified with unique scaling variables. The way in which in which scaling factors are chosen is relevant towards the form of TSM system. The method of components adjustment is described inside the up coming sections. The block diagram on the proposed actual time TSM strategy is shown in Figure 1. All the algorithms utilized in the content analysis block had been described in specifics in earlier papers, as a result they’ll not be talked about here.
The content ana lysis consists selelck kinase inhibitor of. voice exercise detection algorithm, vowel detection algorithm, price of speech estimation, stutter detection and cellphone transitions detection. Since the core in the TSM, a SOLA algorithm was employed. It had been proven that this algorithm ensures large quality of the stretched speech and minimal computational complexity, Additional in excess of, SOLA strategy employs continual values of your analysis time shift and frequent length from the evaluation time frame. This truth permits for integrating the written content evaluation algo rithms together with the TSM method in the natural way, i. e. each time a frame on the input signal is analyzed so as to recognize its content. Subsequently, primarily based on final results presented through the articles examination algorithms, the TSM procedure is performed.
The parameter determin ing the amount of time scale modification is named a scale element, It truly is defined from the equation . the place Sa may be the time shift in the frame utilised through the evaluation stage, Ss could be the time shift on the frame utilised throughout the synthesis phase. In the event the value of is the full details greater than one, the input signal will likely be stretched, if is decrease than one, the signal will likely be shortened. for equal to one, the time scale modification is not going to be carried out. Because the TSM will likely be performed only so that you can expand the time from the input signal, will get values equal or higher than one. Uniform speech stretching In this strategy, a speech signal is stretched applying con stant values on the scaling element. Input signal is time extended only when the voice is detected through the VAD and vowel prolongation was not observed through the vowels detector. Despite the truth that the input signal is non uniformly time scaled, the speech signal is modified uniformly, The stretching procedure is controlled by the d parameter, The value of d really should be specified, Also, elimination of redundancy in the in put signal is performed by changing intervals of silence longer than 200 ms using the time expanded speech.