Latest revision as of 23:31, 26 September 2017

Hearing development in barn owls

Main project page

Previous entry Next entry

General Entries

Played around with filter() and butter() in library.signal MF Kubke 20:11, 28 October 2013 (EDT)

Feedback from Andy

Filter: Higher order is sharper cutoff. A sharp butterworth would be a 5th or 6th order. I wouldn't use any higher.

Try an fft to see where the low pass cutoff should be. If the filter order is too high you get ringing. [ see http://t.co/QfrGaDjvj2 ]

entered wrong sampling rate - sampling rate is actually 50 kHz - need to recalculate the filters based on that. They are off by a factor of 10, so what says lowpassed at 100 is actually at 1k.

Personal Entries

Fabiana

File 2013-10-29-MFK.Rmd

<html>

<body> <h1>Trying to apply a low pass filter to ABR data </h1>

Using package signal

<pre><code class="r">install.packages("signal") </code></pre>

<pre><code>## Error: trying to use CRAN without setting a mirror </code></pre>

<pre><code class="r">library(signal) </code></pre>

<pre><code>## Warning: package 'signal' was built under R version 3.0.2 </code></pre>

<pre><code>## Loading required package: MASS

2. Attaching package: 'signal'
4. The following object is masked from 'package:stats':
6. filter, poly

</code></pre>

Read test file

<pre><code>## [1] "2013-10-27-MFK.Rmd" "2013-10-27.html" "2013-10-27.R"

1. [4] "2013-10-27.txt" "2013-10-28-MFK.html" "2013-10-28-MFK.md"
2. [7] "2013-10-28-MFK.Rmd" "2013-10-29-MFK.Rmd" "Copy189L0A.ABR"
3. [10] "Copy224l0a.txt" "Copy233L0B.ABR.log" "Copy233L0B.ABR.txt"
4. [13] "figure" "knitr-testr.html" "knitr-testr.md"
5. [16] "knitr-testr.Rmd" "kntR2.html" "kntR2.md"
6. [19] "kntR2.Rmd" "my first knitr.html" "my first knitr.md"
7. [22] "my first knitr.Rmd" "mydata_new" "mydata_new.txt"
8. [25] "readrawabr.R" "Sandbox1.R" "Sandbox1.txt"
9. [28] "signal.pdf" "texput.log"

</code></pre>

<pre><code class="r">mydata <- read.table("Copy233L0B.ABR.txt", header = T) head(mydata) #passed test </code></pre>

<pre><code>## msec left right spont bin

1. 1 0.00 1000 1000 1000 1000
2. 2 0.02 1000 1000 1000 1000
3. 3 0.04 1000 1000 1000 1000
4. 4 0.06 1000 1000 1000 1000
5. 5 0.08 1000 1000 1000 1000
6. 6 0.10 1000 1000 1000 1000

</code></pre>

<pre><code class="r">write.table(mydata, "mydata_new.txt") mydata_new <- read.table("mydata_new", header = T) head(mydata_new) #passed test </code></pre>

<pre><code>## msec left right spont bin

1. 1 0.00 1000 1000 1000 1000
2. 2 0.02 1000 1000 1000 1000
3. 3 0.04 1000 1000 1000 1000
4. 4 0.06 1000 1000 1000 1000
5. 5 0.08 1000 1000 1000 1000
6. 6 0.10 1000 1000 1000 1000

</code></pre>

<pre><code class="r">mydata_new$bincomp[1:1000] <- (mydata_new$bin[1:1000] - (mydata_new$left[1:1000] +

   mydata_new$right[1:1000]))

head(mydata_new) </code></pre>

<pre><code>## msec left right spont bin bincomp

1. 1 0.00 1000 1000 1000 1000 -1000
2. 2 0.02 1000 1000 1000 1000 -1000
3. 3 0.04 1000 1000 1000 1000 -1000
4. 4 0.06 1000 1000 1000 1000 -1000
5. 5 0.08 1000 1000 1000 1000 -1000
6. 6 0.10 1000 1000 1000 1000 -1000

</code></pre>

<pre><code class="r">write.table(mydata_new, "mydata_new.txt") #passed test head(mydata_new) </code></pre>

<pre><code>## msec left right spont bin bincomp

1. 1 0.00 1000 1000 1000 1000 -1000
2. 2 0.02 1000 1000 1000 1000 -1000
3. 3 0.04 1000 1000 1000 1000 -1000
4. 4 0.06 1000 1000 1000 1000 -1000
5. 5 0.08 1000 1000 1000 1000 -1000
6. 6 0.10 1000 1000 1000 1000 -1000

</code></pre>

Test plotting for comparison

<pre><code class="r"> plot(mydata_new[c(1, 2)], type = "l", main = "original left") </code></pre>

Now - try to apply a low pass filterusing butter() to determine the properties of the filter.

<pre><code class="r">bf <- butter(3, 0.1) test1 <- filter(bf, mydata_new$left)

par(mfrow = c(1, 2)) plot(mydata_new[c(1, 2)], type = "l", main = "original left") plot(test1, main = "butter(3, 0.1)") </code></pre>

Test type = “low” vs type = “high”

<pre><code class="r">bf <- butter(3, 0.1, type = "low") test1 <- filter(bf, mydata_new$left)

par(mfrow = c(1, 2)) plot(test1, main = "butter type low")

bf <- butter(3, 0.1, type = "high") test1 <- filter(bf, mydata_new$left) plot(test1, main = "butter type high") </code></pre>

(guess default is low then)

Try changing filter properties - change filter order (first argument in function)

<pre><code class="r">par(mfrow = c(2, 2)) plot(mydata_new[c(1, 2)], type = "l", main = "original left")

bf <- butter(1, 0.1, type = "low") test1 <- filter(bf, mydata_new$left) plot(test1, main = "butter(1, 0.1)")

bf <- butter(3, 0.1, type = "low") test1 <- filter(bf, mydata_new$left) plot(test1, main = "butter(3, 0.1)")

bf <- butter(15, 0.1, type = "low") test1 <- filter(bf, mydata_new$left) plot(test1, main = "butter(30, 0.1)") </code></pre>

Not really sure what the order is doing - ????

Stick with filter order == 3, change crirical frequency W. W-> [0-1] where 1 is the Nyquist frequency (½ of the sampling rate )

<pre><code class="r">head(mydata_new) </code></pre>

<pre><code>## msec left right spont bin bincomp

1. 1 0.00 1000 1000 1000 1000 -1000
2. 2 0.02 1000 1000 1000 1000 -1000
3. 3 0.04 1000 1000 1000 1000 -1000
4. 4 0.06 1000 1000 1000 1000 -1000
5. 5 0.08 1000 1000 1000 1000 -1000
6. 6 0.10 1000 1000 1000 1000 -1000

</code></pre>

sampling rate => every 0.2 msec ()

Looks like sampling rate at 5 kHz => Nyquist = 2.5 KHz

<pre><code class="r">par(mfrow = c(2, 2)) plot(mydata_new[c(1, 2)], type = "l", main = "original left")

bf <- butter(3, 0.01, type = "low") test1 <- filter(bf, mydata_new$left) plot(test1, main = "butter(3, 0.01 (25 Hz)")

bf <- butter(3, 0.1, type = "low") test1 <- filter(bf, mydata_new$left) plot(test1, main = "butter(3, 0.1 (250 Hz))")

bf <- butter(3, 1, type = "low") test1 <- filter(bf, mydata_new$left) plot(test1, main = "butter(30, 1 (2.5KHz))") </code></pre>

Try the first derivative on the 250 LP

<pre><code class="r"> bf <- butter(3, 0.1, type = "low") mydata_new$leftlow <- filter(bf, mydata_new$left)

par(mfrow = c(1, 3)) plot(mydata_new$msec, mydata_new$left, type = "l", main = "original left") plot(mydata_new$msec, mydata_new$leftlow, type = "l", main = "lowpassed left") mydata_new$diffleft2[2:1000] <- diff(mydata_new$leftlow)/diff(mydata_new$msec) plot(mydata_new$msec, mydata_new$diffleft2, ylim = c(-3000, 3000), type = "l",

   main = &quot;first diff on lowpassed left&quot;)

</code></pre>

First deriv stil looks noisy - try different filters

<pre><code class="r"> bf <- butter(3, 0.04, type = "low") mydata_new$leftlow <- filter(bf, mydata_new$left)

par(mfrow = c(1, 3)) plot(mydata_new$msec, mydata_new$left, type = "l", main = "original left") plot(mydata_new$msec, mydata_new$leftlow, type = "l", main = "lowpassed 100 left") mydata_new$diffleft2[2:1000] <- diff(mydata_new$leftlow)/diff(mydata_new$msec) plot(mydata_new$msec, mydata_new$diffleft2, ylim = c(-3000, 3000), type = "l",

   main = &quot;first diff on lowpassed left&quot;)

</code></pre>

<pre><code class="r"> par(mfrow = c(1, 1))

plot(mydata_new$msec, mydata_new$diffleft2, ylim = c(-3000, 3000), type = "l",

   main = &quot;first diff on lowpassed left&quot;)

</code></pre>

Mmmmmmm….

</body>

</html>

Andy

Enter content here

Oris

Enter content here

@@ Line 2: / Line 2: @@
 |-
 |style="background-color: #EEE"|[[Image:owwnotebook_icon.png|128px]]<span style="font-size:22px;"> Hearing development in barn owls</span>
-|style="background-color: #F2F2F2" align="center"|<html><img src="/images/9/94/Report.png" border="0" /></html> [[{{#sub:{{FULLPAGENAME}}|0|-11}}|Main project page]]<br />{{#if:{{#lnpreventry:{{FULLPAGENAME}}}}|<html><img src="/images/c/c3/Resultset_previous.png" border="0" /></html>[[{{#lnpreventry:{{FULLPAGENAME}}}}{{!}}Previous entry]]<html>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</html>}}{{#if:{{#lnnextentry:{{FULLPAGENAME}}}}|[[{{#lnnextentry:{{FULLPAGENAME}}}}{{!}}Next entry]]<html><img src="/images/5/5c/Resultset_next.png" border="0" /></html>}}
+|style="background-color: #F2F2F2" align="center"|[[File:Report.png|frameless|link={{#sub:{{FULLPAGENAME}}|0|-11}}]][[{{#sub:{{FULLPAGENAME}}|0|-11}}|Main project page]]<br />{{#if:{{#lnpreventry:{{FULLPAGENAME}}}}|[[File:Resultset_previous.png|frameless|link={{#lnpreventry:{{FULLPAGENAME}}}}]][[{{#lnpreventry:{{FULLPAGENAME}}}}{{!}}Previous entry]]&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;}}{{#if:{{#lnnextentry:{{FULLPAGENAME}}}}|[[{{#lnnextentry:{{FULLPAGENAME}}}}{{!}}Next entry]][[File:Resultset_next.png|frameless|link={{#lnnextentry:{{FULLPAGENAME}}}}]]}}
 |-
 | colspan="2"|
@@ Line 14: / Line 14: @@
 == Feedback from Andy ==
-* Filter: Higher order is sharper cutoff.
+* Filter: Higher order is sharper cutoff. A sharp butterworth would be a 5th or 6th order.  I wouldn't use any higher.
 * Try an fft to see where the low pass cutoff should be. If the filter order is too high you get ringing. [ see http://t.co/QfrGaDjvj2 ]
+* entered wrong sampling rate - sampling rate is actually 50 kHz - need to recalculate the filters based on that. They are off by a factor of 10, so what says lowpassed at 100 is actually at 1k.
 =Personal Entries=

Kubke Lab:Research/ABR/Notebook/2013/10/29: Difference between revisions

Latest revision as of 23:31, 26 September 2017

Contents

General Entries

Feedback from Andy

Personal Entries

Fabiana

Andy

Oris

Navigation menu

Page actions

Page actions

Personal tools

Navigation

Search

research

Tools