The _peak_level function is not robust enough

[XingdaChen1996]

The _peak_level function is not robust enough

When I called the tnr_ecma_st function (_peak_level was called in the function body), I found that changing the length of the first formal parameter "signal" of the tnr_ecma_st function would make tr very unrobust.
Therefore, I changed if Li - spec[temp] < 10 to if Li - spec[temp] < 50. I found that the robustness of tr improved a lot.

[XingdaChen1996]

High-resolution Spectral Analysis(HSA) may also be able to solve this problem. It can greatly improve the resolution and accuracy of single-frequency signals without extending the time signal.

[wantysal]

Hi,

We've already encoutered some issues with this function, there is indeed some conditions lacking to be more robust. HSA could be a great way to overcome this, but there is no existing implementation of it, and the available documentation is not very detailed... If you have some references (articles, codes, ...) concerning the HSA, you can share it here so that we try to work on it ...

We'll try to incorporate a fix in a coming release, I'll keep you updated.
Tks,
Salomé