R/gps.acc.merge.R

In dprocter/modeid: modeid: Identification of travel modes from GPS and accelerometry data

#' @title Merging GPS and accelerometer files
#' @description Returns a data.frame with Accelerometer and GPS data merged by timestamp
#' @param acc.data The accelerometer data to merge, processed with \code{\link{process.acc}}
#' @param gpsfile The GPS file to merge, a .csv
#' @param participant.id
#' A unique identifier for the participant
#' @param epoch.length
#' epoch.length in seconds, currently, 5, 10 or 15 seconds only tested
#' @param british.time
#' boolean, whether or not we the study is in Britain, so we need to check if the data was collected
#' within BST and adjust GPS UTC timings by 1hour
#' @param UTC.offset
#' Hours offset from UTC, e.g. 3 hours ahead, UTC.offset=3, 3 hours behind, UTC=(-3)
#'
#'
#' @details
#' Currently only tested with Qstarz GPS device files. If you need other types contect the author
#' , they can be included with ease
#'
#' @export
gps.acc.merge<-function(acc.data, gpsfile, participant.id,
                        epoch.length, british.time=FALSE, UTC.offset=0){
  
  if (!file.exists(gpsfile)){
    print(paste("no gps file for",participant.id,sep=" "))
  } else{
    
    gps.data<-read.csv(gpsfile)
    
    if (length(gps.data[,1])>0){
      
      gps.data<-subset(gps.data,VALID!=" NO FIX")
      
      if (length(gps.data[,1])>0){
        gps.data$date.time<-paste(gps.data$UTC.DATE,gps.data$UTC.TIME)
        gps.data$date.time<-strptime(gps.data$date.time,format="%Y/%m/%d %H:%M:%S")
        gps.data<-subset(gps.data,UTC.DATE!=" ")
        gps.data$date.time<-gps.data$date.time+(UTC.offset*36000)
        
        # checking whether the file is within BST and adjusting
        if (isTRUE(british.time)){
          BSTstarts<-c("25/03/2012 00:00:00", "31/03/2013 00:00:00","30/03/2014 00:00:00"
                       , "29/03/2015 00:00:00", "27/03/2016 00:00:00", "26/03/2017 00:00:00"
                       , "31/03/2018 00:00:00", "31/03/2019 00:00:00", "29/03/2020 00:00:00"
                       , "28/03/2021 00:00:00" , "27/03/2022 00:00:00")
          BSTends<-c("28/10/2012 00:00:00", "27/10/2013 00:00:00", "26/10/2014 00:00:00"
                     , "25/10/2015 00:00:00", "30/10/2016 00:00:00", "29/10/2017 00:00:00"
                     , "27/10/2018 00:00:00", "27/10/2019 00:00:00", "25/10/2020 00:00:00"
                     , "31/10/2021 00:00:00", "30/10/2022 00:00:00")
          BSTstarts<-strptime(BSTstarts,format="%d/%m/%Y %H:%M:%S")
          BSTends<-strptime(BSTends,format="%d/%m/%Y %H:%M:%S")
          within.st<-0
          
          for (z in 1:length(BSTstarts)){
            if (gps.data$date.time[1]>=BSTstarts[z] & gps.data$date.time[1]<=BSTends[z]){
              within.st<-1
            }
          }
          
          if (within.st==1){
            gps.data$date.time<-gps.data$date.time+3600
          }
          
        } else{
          gps.data$date.time<-paste(gps.data$LOCAL.DATE,gps.data$LOCAL.TIME)
          gps.data$date.time<-strptime(gps.data$date.time,format="%Y/%m/%d %H:%M:%S")
        }
        gps.data$day<-lubridate::wday(gps.data$date.time, label = TRUE, abbr = FALSE)
        
        if (epoch.length==5){
          lubridate::second(gps.data$date.time)<-round(lubridate::second(gps.data$date.time)/5)*5
        }
        # rounding gps data to the nearest epoch
        if (epoch.length==10){
          # round the gps timing to the nearest 10 seconds, so that we can match it with the accelerometry data
          lubridate::second(gps.data$date.time)<-norm.round(lubridate::second(gps.data$date.time),-1)
        }
        
        if (epoch.length==15){
          lubridate::second(gps.data$date.time)<-round(lubridate::second(gps.data$date.time)/15)*15
        }
        
        
        # a variable that will house sum of the signal to noise ratio
        gps.data$sumsnr<-NA
        # satellite info, split into strings
        sat.info<-strsplit(as.character(gps.data$SAT.INFO..SID.ELE.AZI.SNR.),split = ";",fixed=TRUE)
        
        # calculation of sumsnr
        for (i in 1:length(gps.data$INDEX)){
          each.snr<-numeric(length(sat.info[[i]]))
          for (j in 1:length(sat.info[[i]])){
            each.snr[j]<-as.numeric(unlist(strsplit(sat.info[[i]][j],split="-"))[4])
          }
          gps.data$sumsnr[i]<-sum(each.snr)
        }
        
        
        # conversion of the all positive lat and long to +- NS EW
        gps.data$lat<-gps.data$LATITUDE
        gps.data$lat[gps.data$N.S==" S"]<-gps.data$lat[gps.data$N.S==" S"]*-1
        
        gps.data$long<-gps.data$LONGITUDE
        gps.data$long[gps.data$E.W==" W"]<-gps.data$long[gps.data$E.W==" W"]*-1
        
        # getting rid of the units from speed and height
        gps.data$spd<-NA
        gps.data$hi<-NA
        for (i in 1:length(gps.data$INDEX)){
          gps.data$spd[i]<-as.numeric(unlist(strsplit(as.character(gps.data$SPEED[i]),split=" "))[2])
          gps.data$hi[i]<-as.numeric(unlist(strsplit(as.character(gps.data$HEIGHT[i]),split=" "))[2])
        }
        gps.data$date.time.sec<-unclass(as.POSIXct(gps.data$date.time))
        
        
        gps.data<-subset(gps.data,select=c(INDEX,date.time,date.time.sec,day,lat,long,spd,hi,PDOP,HDOP,VDOP,sumsnr))
        names(gps.data)<-c("index","date.time","date.time.sec","day","latitude","longitude","speed","height","pdop","hdop","vdop","sumsnr")
        
        #Stick the two together, kepping all accelerometer data, the missing GPS variables will be NA
        merged.data<-merge(acc.data,gps.data,by="date.time.sec",all.x = TRUE)
        #merged.data<-subset(merged.data,select=-c(date.time.y))
        #names(merged.data)[5]<-"date.time"
        merged.data$date.time<-merged.data$date.time.x
        merged.data<-subset(merged.data,select=-c(date.time.x,date.time.y))
        merged.data$prev.time<-merged.data$date.time
        merged.data$prev.time[1]<-NA
        merged.data$prev.time[2:length(merged.data[,1])]<-merged.data$date.time[1:(length(merged.data[,1])-1)]
        merged.data$remove.dups<-numeric(length(merged.data[,1]))
        merged.data$remove.dups[merged.data$date.time==merged.data$prev.time]<-1
        merged.data<-subset(merged.data,remove.dups!=1)
        merged.data<-subset(merged.data,select=-c(remove.dups,prev.time,date.time.sec))
        
        merged.data$acceleration<-NA
        for (j in 2:length(merged.data[,1])){
          if (!is.na(merged.data$speed[j]) & !is.na(merged.data$speed[j-1])){
            merged.data$acceleration[j]<-(merged.data$speed[j]-merged.data$speed[j-1])/epoch.length
          }
        }
        
        return(merged.data)
      } else{
        print(paste(participant.id, "has no GPS data, ignoring", sep=" "))
      }
    } else{
      print(paste(participant.id, "has no GPS data, ignoring", sep=" "))
    }
  }
}