In his article: "Playing TRIX: The Triple Exponential Smoothing Oscillator" Joe Luisi argues that using the linear regression value as a signal line for a TRIX is better than using a slope.
Without attempting to prove or disapprove is that is correct, here is the version that does that. The signals are not bad but some experimenting with parameters is advised.
Corrected T3STD Trend Envelopes of Averages - multi time frame
STD Trend Envelopes of Averages - multi time frame