Waveform　selection　plays　an　important　role　in　the　processing　of　in-line　terahertz　measurements　of　pharmaceutical　tablet　coating　processes.　This　paper　presents　an　approach　to　optimise　waveform　selection　by　utilising　an　artificial　recurrent　neural　network　and　transfer　learning.　The　results　show　that　the　averaged　coating　thickness　gradually　increases　throughout　the　coating　process.　In　comparison　with　the　conventional　method,　our　approach　allows　more　than　double　the　number　of　waveforms　to　be　selected　without　compromising　on　the　accuracy　when　compared　against　off-line　measurements.　Moreover,　the　processing　time　of　waveform　selection　decreases　so　that　it　can　be　applied　for　real-time　coating　monitor　in　the　pharmaceutical　industry.　©　2021　IEEE.