Open GoogleCodeExporter opened 9 years ago
great to see voice navigation on monav! thanks!
1. the instructions on screen seemed correct, but voice directions didn't agree
with them. are you taking another gps reading for the voice instructions rather
than sharing the gps reading with the written instructions? would probably be
good to 'share' the same data.
2. the phantom 'turn right' instructions could be due to monav being too quick
to decide that you had gone 'off route' due to a bad gps reading. i logged a
bug with the marble project saying that marble was a little slow to detect you
had gone off route, but now i think they made the correct decision as
inaccurate gps readings can have you in the wrong lane/phantom right turns etc.
so monav should take a leaf out of marble's book and not be so quick to decide
youve gone off route.
3. i might log a feature request with monav for the route colo(u)r , as it was
impossible to see the blue route on the blue road.
4. i guess the biggest thing missing from this great prototype is advance
warning of the next turning direction. i.e. adding a "THEN" to instructions.
'continue for x meters THEN turn right'
looks very good for a protype though!!
Original comment by bugspamm...@yahoo.com
on 25 Aug 2011 at 7:45
1) What do you mean they don't agree? I didn't record "Continue for X meters"
for all values of X. Currently it underestimates the distance. Can you give
some specific position in the screencast that shows the issue you mean?
2) Yes I realized the cause myself too. I think this is the largest
show-stopper now.
3) Sure, why not. I'm mainly interested in being able to drive without having
to look at the display.
4) This is easy to implement.
Original comment by timo.lin...@gmail.com
on 25 Aug 2011 at 8:15
Do you have any opinion on how the actual audio output part should be done? I
googled around and it seems that Phonon might be the right API to use. My
prototype just uses
system("aplay " + audioCommand[0].toAscii().data() )
in MainWindow::instructionsChanged at the moment. (This is obviously wrong for
many reasons.)
Original comment by timo.lin...@gmail.com
on 25 Aug 2011 at 8:22
theres a bug with phonon whereby it doesn't do gapless playback on some mobile
platforms (e.g. nokia n900) according to the marble guys. so you get 'in
'...pause.. '50 meters'...pause.. 'turn right' with each pause being about 1
second. so that's something to be aware of.. the marble guys took advance (i.e.
in .. x meters turn right)voice instructions out of the n900 version as a
result. so now the n900 just has the 'turn right' instruction when you actually
reach the junction. the marble guys *may* have logged a phonon bug for this.
you could maybe get around this by pre-joining the sound files somehow.
as for screen instructions not matching voice instructions, there's a point
there the screen says '8km' and the voice says '5km' and you also write about
it in the notepad on the right if i remember correctly.
Original comment by bugspamm...@yahoo.com
on 26 Aug 2011 at 12:41
Yes I recorded only 5km and 10km so at 8km it read 5km :-)
Original comment by timo.lin...@gmail.com
on 26 Aug 2011 at 9:12
Sweet.
Would you be interested in pursuing this further?
Original comment by veaac.fd...@gmail.com
on 28 Aug 2011 at 10:40
Definitely. Monav is great but there are two show-stopper problems for me
1) no speech output (this ticket #73)
2) it doesn't always find the streets even though I have added them to
openstreetmap (ticket #62)
For #73 the largest problems are
1.1) monav needs to cope better with inaccurate GPS. I don't know how to do
that.
1.2) monav sometimes decides to route me away from a highway and then back.
This causes really confusing "Take the ramp" instructions.
1.3) I need to figure out how to use phonon. The prototype just uses external
command since I only wanted to see what other parts of monav need to be
improved for speech output to make sense. I'm attaching the patch that I used
in the screencast so that you can get ideas. The audio files are in
http://iki.fi/lindi/monav/audio.tar.gz
Ticket #62 is probably trivial once you find out which part of the preprocessor
decides to demolish my street ;-)
Original comment by timo.lin...@gmail.com
on 28 Aug 2011 at 11:27
Attachments:
Original issue reported on code.google.com by
timo.lin...@gmail.com
on 25 Aug 2011 at 10:07