Ifbyphone Blog

Many new telephone services, such as Click to Call, Virtual Call Centers, Hosted IVR and Voice Broadcast, rely on voice over IP (VOIP) telephone connections to deliver services at the lowest possible cost. While some of these services sound great, others deliver voice quality no better than the walkie talkies your child recently received for Christmas. This blog entry will attempt to present, for a non technical audience, the basics behind VOIP communications that determine the quality of a phone call. Specifically this blog will address VOIP traffic transmitted over the public Internet.

Your voice starts out as an analog wave or vibration. Remember the telephone you built as a child with two cans and a string. The bottom of the first can vibrates, this vibrates the string, which then vibrates the bottom of the second can. In a traditional telephone these same vibrations are used to modulate an electrical signal transmitted over a wire which replaces the string.In VOIP communications your voice is converted from an analog wave into a series of data packets. These data packets, which share the Internet with email, web page transmissions and file transfers, are then transmitted from one computer to another where at the destination the data packets are converted back into a sound wave you can hear.

The process of digitizing sound is well understand and as in the case of DVDs can be very high quality. In fact many DVDs are better quality than traditional analog recordings.Unfortunately, unlike high quality DVDs, which have the capacity to transport a large amount of data, the Internet has limited bandwidth. Therefore when we digitize voice for transmission over the Internet we compress the data significantly more than the music on a typical DVD. This compression reduces the quality of the sound.

There are many standards for the compression of voice. However, in a typical engineering trade off; the more we compress the voice transmission, the more conversations we can transmit over the same Internet connection. The other side of this trade off is that the more we compress the voice the worse it sounds.

The vast majority of telephone calls today, even the calls made on your traditional home phone, are digitized. The traditional telephone companies use a compression standard called PCM or G.711. In fact VOIP telephone services that use G.711, such as Ifbyphone, sound just as good as traditional telephone calls.

Another common compression standard is G.729. Telephone calls compressed with G.729 are compressed on average four times more than G.711 or traditional telephone calls. This additional compression results in lower overall call quality.The telephone industry has a standard measurement of telephone call voice quality called the Mean Opinion Score or MOS. The generally accepted typical MOS (just Google MOS for more examples) for a number of compression standards follow:

Traditional PSTN phone calls: MOS is close to 4.5

G.711: 4.4 or better

G.729 about 4

G.723 about 3.5

G.726 between 3.5 and 4.3

Unfortunately this is not the whole picture. The quality of the connection of to the Internet used by your VOIP service provider also affects the quality of your telephone call. And finally some VOIP vendors use a Voice Activity Detection algorithm to reduce the bandwidth they need for your calls. VAD attempts to listen for silence between the callers in a telephone conversation and stop transmitting packages when no one is talking. If the VAD algorithm is inappropriately tuned the ends of your words will be clipped and overall voice quality will decline.

At Ifbyphone we are committed to the delivery of high quality voice services designed for business customers. We exclusively use G.711 for our business services. Since many of our services rely on automated voice recognition we need to set very high voice standards for the recognition to function properly.

Please feel free to post comments to this blog entry about your VOIP experiences and information you may have about the codex or compression standard used by your VOIP vendor.

Need an easy, attractive button to front a Click-to-call link or another page?  My cool button is a free Beta site that allows you to customize simple Web 2.0 buttons with an array of colors and icons, and will let you enter in your own text.  If you’re using an embedded Click-to-call, design the “Call” button here, or place text on a site that says “To speak with us now” and make the button below say “Click here!”  The possibilities are endless (well, almost – we’ll see what this nifty site does with future Beta releases).

In a video interview posted on TechRepublic, Bill Gates discusses some of the ways that the standard telephone sitting on your desk with too many buttons and a tiny screen will change in the coming years.  Only one in three people know how to successfully transfer a phone call to someone else in their office, according to Gates, who says that the desk phone has hardly evolved at all in recent times.  Compared to a mobile phone or a Mac, he’s right – remember those four-inch floppy disks you used to use to save your Clarisworks files?  In the next five years, Gates sees the phone evolving into an extension of the PC or a stand-alone phone with a larger touch screen, scrolling through transfer names, and other features foreign to the common desk phone.  Speech recognition is in his crystal ball as well.  Take a look at the video.

One of the most important things I do when I write an application is check for consistent prompts. What do we say to the customer? Do we say the same thing everywhere, or do we suddenly change our minds half-way through the application?Here’s an example from what I’m working on today. The client supplied the following prompt:

Press # or say “main menu” at any time during this message to go to the options menu.

How do I check this prompt to make certain that it’s (a) consistent and therefore (b) not confusing?I start out by underlining each individual “name” and each individual “action” in the prompt. A “name” is the name of something: for example, what you and I would call an account number some applications call a “customer number.” An “action” is something that we ask the customer to do, or that the customer wants to do; for example, we might ask the customer to “say” the account number, “press” a button, or the customer may wish to “transfer” funds.Here’s a table of all the names and actions I found in the prompt, in order:

Then I start checking for consistency. I looked through this application and made a decision: during the application, we will always use “press” consistently; we won’t use “press” in some places and “enter” in others. We will also use “say” consistently.But the rest of this prompt poses further problems. First of all, the action “#” (pressing that particular key) isn’t consistent with other places; it’s the only # in the entire application. The “options menu” wasn’t defined anywhere — in fact, although I am developing the software, even I can’t figure out which menu is the “options” menu. And furthermore why require someone to say “main menu” if they want to go to the “options” menu — that’s not consistent even in the same prompt! Actions should always be consistent with names whenever possible.”Go to” is an action that the user might want to do, but the wording has to be consistent with other places in the application. Finally, the name “this message” might not be consistent with the idea that we’re in a “menu.” Is this a “message” or a “menu?”In the end, I was able to straighten this out. The call flow is really just returning the user to the previous menu, and now I use the words “previous menu” everywhere in the application and also always use the * key to or the words “previous menu” to get there. I also use “return to” when I’m returning rather than “go to.” The prompt now reads something like this:

To return to the previous menu, say “previous menu” or press *.

I use the exact same wording for this prompt everywhere in the document.With consistent prompts, callers will be more comfortable with the application because they won’t have to re-learn how the application behaves each time they encounter a prompt, and they can focus on the content of the questions rather than trying to figure out what the questions mean in the first place.

A recent cover article in Speech Technology Magazine discusses small business and the adoption of speech technology. Of course, what they mean by “small” is one thousand employees or less; these businesses are just now finding that the purchase of special-purpose speech technology solutions and equipment can be cost effective.

I don’t know the size of a typical Ifbyphone customer; but Ifbyphone’s technology drives the cost of speech recognition down to the point that a single-person company can find it useful, affordable, and indispensable.