Overview of Speech Experiment
This study is meant to collect speech utterances: instances of people saying the same sentence in a normal and in a whispered voice. The collected data will be used to test and improve speech recognition algorithms.
What a typical session will be like
Our sessions will go something like this:
- I will first explain what this study is about, and what you have to do.
- You will need to sign a consent form, agreeing to let us record your speech.
- A copy of the consent form will be made available to you upon request
- You will finish speaking and whispering all sentences.
- If we cannot finish everything in our alloted time, we will reschedule a time.
- You will sign a receipt of payment, and I will pay you for your time.
There are nine sets of sentences, each set should take less than ten minutes to both read and whisper. You should not spend more than two hours finishing the whole experiment.
Instructions For Producing Good Data
To speed up the process, and not waste your time, you're advised to pay attention to the following:
- Speak fluently and articulately.
- Maintain a natural, constant and comfortable volume.
- If you are not sure how to pronounce a word, guess. It does not matter so long as you are consistent
- Don't pause unnecessarily
- Don't stutter, or start and stop on words.
- Don't repeat words
- It is probably ok if you miss a word or two, so long as it is not too far off.
- Please take breaks when you are tired. Take regular sips of water.
- Water will be provided.
Venue and Scheduling
We will do our recording in the anechoic chamber in RM G90 Phonetics Lab, in the Foreign Language Building . This is in the basement of FLB.
The anechoic chamber is a limited resource shared among all students in the linguistic department, so we will have to schedule in advance to ensure that we get to use it. We are only allowed two-hour bookings at a maximum, unless nobody else is using the chamber.
I should be free anytime during the week to do the recordings. Please email me to work around our schedules.
If we cannot finish due to technical fault, we will need to reschedule. You will be paid for my mistakes =P
In case we do not finish, we will need to reschedule.
For more information
Here are some additional materials:
This study is funded by the Campus Research Board, and approved by the Institutional Review Board
Our sentences are taken from the TIMIT phonetically balanced set. There should not be any words that seem strange to you.
Please direct any queries to Boon Pang, Lim.
Tue Oct 9 22:39:49 EDT 2007