Table of Group Meeting
| Date | Activity
|
| 12 Aug:
| Form group, 4 members Roger, Mark, Matthew and Jia
|
| 19 Aug:
| Topic is not yet release.......
|
| 21 Aug:
| Speaker Recognition.... sounds interesting, since all member agree on this topic, then let's go... project of 60%
|
| 28 Aug:
| Starting to research about all the infomation in the topic of Speaker recognition and permanently split the duties to each of the member
|
| 31 Aug:
| All memeber come up with serveral methods to approch the the task: after filtering, our option are: Spectrgram, VQ, power density function, FFT
|
| 5 Sep:
| each memeber come up with a proposal of each method, and did a mini presentation to eachothers in the group.
|
| 6 Sep:
| we having a polling on the day, to select a approiate method, finally we decide to use the FFT and Spectrgram. ok let's start work.
|
| 13 Sep:
| we construct a first draft layout of our Wiki documentation, and post some thoeries about the topic: speaker recognition.
|
| 20 Sep:
| plan to start our coding, but we confuse about the requirement of the "two mode", training and recognition. speak to the tutor, they say ignore them. fly away
|
| 27 Sep:
| After 1 week of data collection, we currently got 10 voices on hand, and we apply the Matlab function "Spectrgram" to them, we analysis the difference between male and female voice and try to outline some characterstics, plus more search on the thoery.
|
| 4 Oct:
| Start our view from the frequency perspective of a voice, apply each voice using Matlab FFT function, by now we got 19 voices all together.
|
| 11 Oct:
| By apply all voices to the FFT function, we concluded some of the characterstics between male and female voice. our team decide to use energy ratio as a core element in our system, use to distinguish between 2 voices, however we need a way to calucalte the energy ratio, from those approachs we start to design Matlab filters to fit our needs.
|
| 18 Oct:
| After we propose our ideas to the tutor, we gain alots of the feedback, we realise we need more than 1 filter to fully analysis our samples, we need to design 3 filters: lowpass, highpass and bandpass. all 3 filters are design to fit the male voice's characterstics, and we set a counter in our code, if the input voice match with the filter the counter increment by 1, any male voice should have a counter value of 2 or more.
|
| 25 Oct:
| 1 week to go til the code is due. we currently on the final stage of editing and implement on our Matlab code, all code seems to work all-right and we quite happy with the system output, we gain a accuracy up to almost 90%.
|
| 1 Nov:
| Due date!!!! last chance to finalise our code, adding comments and explaination in each main function, and link our system to the GUI we had implement few days ago, the GUI seems to be simple and neat, however it do it's job. burn it into a cd and that is! DUE
|
| 6 Nov:
| Wiki documentation is due this wednesday, each memebr is allocated a section to complete.......
|
