Welch & Hilbert

Hello IgorList,

I am interested in estimating the “amount of motion” in a 14-second series of angular velocity values. They were obtained from an inertial sensor attached to the wrist as the hand made a pronation-supination motion (as if screwing in a light bulb). The time series is approximately sinusoidal. I’d like to go a step beyond a simple FFT power spectrum, and I wonder if anyone has written code that they’d like to share for calculating the following:

1. FFT with Welch’s method (divides a finite time series into smaller chunks and partially overlaps them, as a way to reduce the bias in favor of end regions due to the finite nature of the series).

2. Analytic signal: extract the envelope of an oscillating signal. I think Hilbert had something to do with this but my background is too limited to know whether this is the same thing as the Hilbert transform. In my case I am interested in estimating the peak-to-peak magnitude over time, while ignoring small irregularities between the large peaks.

Best regards,

Pietro
1. DSPPeriodogram with /SEGN


2. HilbertTransform. See the Example.

--Jim Prouty
Software Engineer, WaveMetrics, Inc.
Sorry for my delayed reply. I was under the impression that I was getting daily digests of the posts in this forum through the Igor mailing list.

Thanks for pointing me to the /SEGN flag for DSPperiodogram

The HilbertTransform example does exactly what I want to do.

Pietro