Implement speech synthesis in ASP.NET program

Author：Eve Cole Update Time：2009-06-30 16:07:51

You may have noticed that speech synthesis is becoming more and more common these days. Hundreds of electronic toys and gadgets are capable of speaking to their owners in machine voices. Whether it's a talking baby doll, a talking pedometer or a new automated phone system, a product is released almost every day that uses text to speak.

Text-to-speech (TTS) technology, also known as speech synthesis, is the process of converting input text into audio speech. This approach is better than pre-recorded text - it requires advance knowledge of what is to be said. With TTS, it is possible to add dynamic information. The dynamic information can come from a database; or it can be repeated by the user.

1. TTS test

If you've seen or heard of TTS in action, then you might want to download a free version of ReadPlease 2003. This product is able to read text from the Windows clipboard. To use it, you simply paste some text into the ReadPlease editor (see Figure 1) - and assuming your PC speakers are turned on, you'll hear the text play back. Currently, the product only works on all Windows desktop OS versions, but they also plan to release versions for Mac, Unix, Palm, and Windows CE systems.

The interesting thing about ReadPlease is that you can use the ReadPlease editor to experiment with your TTS parameter selections. For example, you can adjust the articulation speed by moving the Speed slider control shown in Figure 1 up or down. You can also change the voice used by clicking the arrow button below the little face icon.

Figure 1. ReadPlease 2003 application: This Windows application may read any text pasted from the clipboard into an edit field.

What do you need? Visual Studio.NET 2003, Microsoft Speech Applications SDK 1.1

Clicking the "Tools" menu and then selecting "Options" allows you to further experiment with the TTS editor; for example, you can adjust the length of the speech engine's pauses between paragraphs. Upgrading to the ReadPlease Plus version will give you access to an additional pronunciation editor option that lets you specify how to pronounce a particular word. ReadPlease Plus also includes a taskbar that you can dock at the top of your Windows desktop. So you can quickly drag text from any file-based application onto the taskbar and have it read to you.

By default, the ReadPlease application uses the built-in Microsoft voice (Mark, Mike, Sam or Marilyn). However, you can optionally purchase higher-quality AT&T Natural Voices, such as those available in the AT&T Natural Voices Starter Pack. The Starter Pack includes 8K versions of "Mike" and "Crystal," both of which sound better than the default Microsoft voices.

Author's Note: The term 8K means that the sampling frequency used to create WAV files occurs at 8000 bits per second. Another format is 16k, which produces a clearer and more natural listening effect. Basically, the higher the sample rate, the better the sound quality.

2. Microsoft Voice Application Development Kit SDK 1.1

In 2004, Microsoft released Microsoft Speech Server, and there is a free SDK that allows you to develop Web-based speech programs that run on Speech Server. You can use the SDK to build telephony or voice-only applications - where computer-to-user interaction is achieved using the phone. You can also build multimodal applications where the user chooses to use voice or traditional web controls as input.

The Microsoft TTS engine synthesizes text by splitting words into phonemes. Phonemes are the basic units of human language. They represent a group of "sounds" that make up the sounds of a word. The extracted phonemes are then analyzed by the TTS engine and converted into symbols used to generate digital audio speech.

You can use the downloadable sample application that accompanies this article (ExploringTextToSpeech.csproj) to experiment with the configurable aspects of the Microsoft TTS engine. This multimodal application consists of a web page (see Figure 2) into which you can enter some text. You can then click a button to listen to the text in the following manner.

Author's Note: Using a TTS engine is unavoidable when the text to be played cannot be known in advance; however, you can get better sound quality by recording audio. When audio quality is critical, you can use the Microsoft Speech Application Software Development Kit (SASDK) to record audio. For example, you might want to use recorded audio to prompt the user with information. The recorded audio can be broken down into a series of cues, which are then concatenated at run time.

· Speak Text Normally-Provides a test benchmark

· Say as an Acronym - The text "ASP" is generated by "ASP"

· Say as Name-Mr.John Doe was attacked "Mister John Doe"

· Say As Date - In this case, the date is formatted as: month, day, year

· Say as Web Address - In this case, the text is formatted as a Uniform Resource Identifier (URI)

· Say as Digits - Numbers entered as text are represented as a series of digits

· High Pitch/Slow Rate - The text is emitted slowly and in a high pitch.

· Rate Fast/Volume Loud - Text is emitted at a fast high pitch.

· Low Pitch/Volume Soft-The text is emitted in a low pitch at a moderate volume.

Figure 2. Sample application - You can use this application to listen to text spoken in a variety of ways.

The multimodal application uses a prompt control to specify the audio that will be played to the user. The prompt control contains the InlineContent property - it can contain a basic Speech control that is either Content or Value. The Content control specifies a specific prompt file containing the stored audio recording. The Value control specifies an element in an HTML Web page. The sample application uses a Value control that references an input element named txtText (in the "Type some text here:" field in Figure 2). Here is the HTML describing the annotation used for a prompt:

＜speech:prompt id="prmText" runat="server"＞
＜InlineContent＞
＜speech:Value runat="server"
TargetElement="txtText"
TargetAttribute="value">
＜/speech:Value＞
＜/InlineContent＞
＜/speech:prompt＞

3. Speech synthesis annotation language

The TTS engine makes certain default assumptions about how to play text referenced by the InlineContent attribute, but developers can control how the TTS engine generates audio by using Speech Synthesis Markup Language (SSML) elements. According to the recommendation of the W3C Association, SSML is an XML-based markup language. Table 1 lists the SSML elements supported by SASDK.

Table 1. Supported SSML elements: The table lists the SSML elements supported by the SASDK and used to control the way the TTS engine generates text.

SSML element	description
ssml:paragraph/ssml:sentence	is used to break text into sentences or paragraphs.
ssml:say-as	is used to specify how to play text. It accepts several different attributes to specify the type of text.
ssml:phoneme	is used to control the pronunciation of a word.
ssml:sub	is used to specify a replacement word or phrase to replace the specified text.
ssml:emphasis	is used to increase the stress placed on a word or phrase.
ssml:break	is used to insert some pauses between certain words in the text.
ssml:prosody	is used to control pitch, rate and volume.
ssml:audio	is used to insert recorded audio files.
ssml:mark	is used to insert a mark at a certain position in the text. This tag can then be used to represent an event or trigger a behavior.

This sample application illustrates the use of say-as and prosodySSML elements. Each button on the Default.aspx page corresponds to a prompt control. These prompt controls include an ssml:say-as or an ssml:prosody element within the InlineContent element. The following example shows the HTML markup for one of these elements:

＜speech:prompt id="prmSayAsAcronym" runat="server"＞
＜InlineContent＞
＜ssml:say-as type="acronym"＞
＜speech:Value runat="server"
TargetElement="txtText"
TargetAttribute="value"></speech:Value>
＜/ssml:say-as＞
＜/InlineContent＞
＜/speech:prompt＞

When the user clicks one of these buttons, it executes the JavaScript shown below:

function SayAsAcronym() {
prmSayAsAcronym.Start();
}

In the example above, the prompt named prmSayAsAcronym includes the ssml:say-as element, which specifies that any text contained in a txtText input element should be emitted with the first letter of each word. So, if you type "ASP" into a text element and click "Say As Acronym", the TTS engine will read each letter.

To experiment with the sample application, you can enter some text snippets and click each button to see how the TTS engine interprets the text. I advocate that you change element values and experiment with the way each control is generated. The SASDK gives developers good control over how the TTS engine generates text, so experimentation may lead to a more natural speech-based application.