Voiceprint Lock

Last time I talked about some applications about speaker recognition. In this blog, I will mainly focus on one of them, which is the voiceprint lock.2a

In my point of view, the most difficult part in speaker recognition is when you use it in noise environment. The noise from environment will be interference of the real useful signal. They are not many voiceprint lock software in China till now. But as far as I know, it is widely used in countries like Japan and USA.

The mostly used Chinese voiceprint software are LingXivoice and SIVI. Unfortunately, they only works on android, and maybe they are now considering going to iOS market. Those software are easy to use, similar to set a password for your phone. They are text-dependent, which can give more accurate result. Text-independent recognition would be an ideal case, but what you are saying also depends on your mood and your health condition, for instance, catch a cold may change your voice a bit. Besides, for a text-independent recognition, it may also detect the voice of the owner’s recordings, which is not safe for the product.

When you set a voiceprint lock, in most cases it first give you different texts to record, each text will be recorded three times. The time when you want to unlock it, it gives you a random text which is already be stored when your set the voiceprint lock, then you can unlock your phone if verified.

That is the introduction to voiceprint lock application. If you have an android device, you can experience it by yourself now!

Source:

http://research.microsoft.com/enus/um/people/zhang/speaker%20verification/default.htm

http://game.3533.com/ruanjian/4730.htm

Advertisements

5 comments

  1. samwendrickx · December 4, 2014

    Hi,

    I was wondering in what field you imagine your thesis work will be used when it’s done.
    Are you designing it for a voice lock like in this article? If so how would you be checking for attempts at counterfit?
    Would you be able to recognize a recording from someone’s voice?
    Also, you mentioned in your previous post that fingerprinting may be difficult when your finger is hurt. I imagine having a sore throat or a cold could do the same for a voice lock. Are there back-up plans for this?

    Kind regards

    Sam

    Like

    • wumiaomiao201410 · December 6, 2014

      Thanks for your comment.
      Actually I don’t know till now how the company want to use it when our work is done. I’ll reply you when I get it in the future.
      When having a sore throat or a cold, I think probably the recognition won’t work。

      Like

    • Duowei · February 28, 2015

      Speaker recognition is a very big topic, we will only do a very tiny step in our thesis. We may not create a product that can identify a speaker with no condition required or can use for all the applications. What we will do is to analysis the results between different algorithms, and come up with some results as a guideline for the future developers.
      For the question about if you get a sore throat of cold, one solution will be you record both your normal voice and voice when you get cold in database. Just like you record different finger prints. But it will be silly and inconvenient. They smart way is maybe using some algorithms to cancel out the channel effects(it means all the changes between the source and receiver). It is still advanced for us, we will learn that more after finishing the main topics. But it is fun to think about it!

      Duowei

      Like

  2. Sam B. · December 8, 2014

    Hello!

    Passwords that you can reproduce by typing or other forms of physical manipulation are fairly time independent passwords. However, like you mentioned in your post, if you become ill and your voice changes as a result you are going to be locked out of your system as a result for a few days. This in mind, what do you think they suggest for pubescent teenagers, primarily boys. Their voices will be fundamentally different over the course of a few years. Does this need special algorithms or do they need to make new recordings periodically when their voice still sounds close enough to their old samples to log in, but different enough to cause them to worry about their future access?

    Sam B.

    Like

    • wumiaomiao201410 · December 8, 2014

      Thanks for your comment, sam!
      I think they need to make new recordings as their voice changes.
      If you are using a phone with fingerprint, You’ll find their back up plan is still password lock. The sam can be with voiceprint lock, so they don’t need to worry that much.

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s