process.sh

#!/bin/bash

# Filters sounds and schedules posts

# Project home directory
homedir=/home/pi/barkdetect/

# Classify sounds, writes class to ${homedir}classification.txt
python $homedir/pyAudioAnalysis/audioAnalysis.py classifyFolder -i ${homedir}samples --model knn --classifier ${homedir}knnBarkDetect --details > ${homedir}classification.txt

#Check to see if classification ran correctly
if [[ $? == 0 ]]; then

	# Iterate over unfiltered samples
	for filename in ${homedir}samples/snd*.wav
	do
	    # Find the classification of the current file
	    res=`grep $filename < ${homedir}classification.txt`

	    # Filter into appropriate file
	    if [[ $res == *"barks"* ]]; then
		mv $filename ${homedir}samples/barks
	    elif [[ $res == *"junk"* ]]; then
		mv $filename ${homedir}samples/junk
	    fi
	done

	# Check that unprocessed barks directory is not empty to avoid pointless error message
	if [[ `ls -p ${homedir}samples/barks | grep -v /` ]]; then
		# Schedule a twitter post for each filtered bark, 
		# delayed 10 minutes from original recording
		for file in ${homedir}samples/barks/snd*.wav; do
		    at -f ${homedir}atjob.txt `${homedir}getTimeFromFilename.py $file`
		    mv $file ${homedir}samples/barks/processed
		done
	fi

# if classification did not complete correctly, clear the unprocessed samples, probably a corrupted file.
else
	rm ${homedir}samples/snd*.wav
fi