pzl / dogwatch Goto Github PK

getting:

*** Error in `./dog2raw': free(): invalid next size (normal): 0x000000000081f800 ***

on a rather large (3h) .dog file

snprintf or sprintf_s if possible

http://natashenka.ca/sprintf/

This includes the sound data as well as any application data (identified barks, date, etc).

Currently already using the lowest fidelity settings possible (8KHz, mono, 8bit). But still uses a significant filesize (60s * 8000/s * 1ch * 8bits) = 468 kb per minute (or 27MB/hr).

Sure, 137MB for 5 hours of recording is not bad, but it's potentially a lot of useless data. For times when the house is quiet, that's MBs of silence.

Consider a compressed format for file-writing or storage.

Are noisy cats a thing? What are other pets you have in a house that might make noise?

BIRDS.

turtles
fish
lizards
snakes

Pets that might not emit noises but could cause a ruckus in their environment (cage or house given free roam)

• guinea pig, hamsters, small rodents
• ferret ?
• rabbit (cage biting, thumping)

In most cases, a dog's bark should be louder than the ambient noise at rest. Basic bark detection? Watch for peaks

use ML to learn the particulars of a home's ambient background noise, relative bark volume and pitch of their dog.

Can try passively training to filter out background noise during quiet periods. When fairly confident that no barking is occurring, characterize the present noise to filter out continuously.

Can also actively train on live data by allowing to user to manually re-characterize barks as "This is not a bark" or partially-confident noises as "yes/no". Should have easy sound-clip playback access for noises of interest

support command line options

.dog now contains mixed control sequences and PCM data. Need a clean (library) way of reading PCM data from file for processing e.g.

SAMPLE *read_dogfile(int bytes);

needs to handle boundaries between compression expansion

No longer using sox/arecord. Update README for new program, usage, etc

Make an end-user friendly into page to this project, aimed at downloading and using

Employ more advanced DSP like FFT or Goertzel algorithm to help positively identify barks.