johnnyzhangzhao / monav

Automatically exported from code.google.com/p/monav
0 stars 0 forks source link

spoken driving instructions #73

Open GoogleCodeExporter opened 8 years ago

GoogleCodeExporter commented 8 years ago
I did a small prototype that reads spoken routing instructions.

Please take a look at the screencast at

http://lindi.iki.fi/lindi/screencast/monav-speech-routing-experiment1.ogv

Original issue reported on code.google.com by timo.lin...@gmail.com on 25 Aug 2011 at 10:07

GoogleCodeExporter commented 8 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

GoogleCodeExporter commented 8 years ago
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

GoogleCodeExporter commented 8 years ago
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

GoogleCodeExporter commented 8 years ago
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

GoogleCodeExporter commented 8 years ago
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

GoogleCodeExporter commented 8 years ago
Sweet.

Would you be interested in pursuing this further?

Original comment by veaac.fd...@gmail.com on 28 Aug 2011 at 10:40

GoogleCodeExporter commented 8 years ago
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: