Word Recognition is the basic mode in speech recognition technology.  It exist several default words in the system. When the user speaks, the voice is conducted by signal processing, features are extracted and similarities calculated, and then the highest score among the words result in recognizing the correct word. Word Recognition uses the full match method to recognize what the user says.

In Keyword Spotting, the user can speak any combination of words and the keywords are recognized while filler words are filtered out. Keyword Spotting can extract and recognize certain keywords from the user's speech.

Auto dialing system has been widely used in call center all over the world. Most of dialing systems transfer the calls to agents after the calls are connected; however, for the calls were picked up by answer machine or voice mail are not necessary to be transferred to agents. Besides call transfer, redial is a convenient way to reach the numbers which have not been connected; however, it works worthless if the system repeating dial the numbers are invalid, suspended, or non-existed. Through Delta Signal Detector, the most innovated and intelligent system, the call respondents can be easily distinguished and categorized to 14 groups which shown as below. By sorting these 14 groups, dialing system can be programmable to route the calls in proper direction and works more efficiently in call center.

14 Groups:
- Voice mail box
- Turn off
- No answer
- Can not be connected
- In Suspension
- Non-existed phone number
- In suspension or no response
- No answer and turn to voice mail box
- Busy
- Busy  with call waiting
- Busy and turn to voicemail box
- Phone number is not registered
- Phone number is  not acceptable
- Do not disturb

One of smart designs in Delta Signal Detector is to delete invalid numbers from database automatically if the numbers had been detected as invalid number by system, or to leave a short message in mailbox if the calls are answered by voice mail.

The signal detection  product provides C/C++ API functions for  application developer use. The application can implements streaming recording and feeds-in the streaming data to the signal detection API in order to  quickly get the call result in a real-time response.

Delta Speaker Verification can be divided into two types of technologies. One is related to a text-dependent test, that is, the voice content of individual voice print authentication database and the testing voice must be the same. The other is related to the text-independent test, that is, the voice content of an individual voice print authentication database and the testing voice are allowed to be different. We provide both types mentioned above to allow users a wider application.

Speaker verification technology is widely used. Personal aspects can be applied to control equipment, such as user permission for personal computers. For enterprise applications, speaker verification can be used as fraud detection for credit card customers. It can also be used to extract specific speakers’ call logs from a large number of customer service dialogues, which would reduce the workload of customer service agents and improve customer service quality (Quality Management).

Text to Speech is the technology  to  transfer text to speech, and it’s also called as Speech Synthesis. Input a text, processed by the following module, and then generate the output speech.

- Text Processing Module: Extract synthesis information from text content
- Prosody Generating Module: estimate prosody from text information
- Signal Processing Module: adjust speech data by prosody

The word separated by space is small paragraph，comma is middle paragraph, and period is large paragraph; then finding corresponding syllable from speech corpus, adjusted by proper prosody, and generate the output speech.

Delta focuses on Syllable base Text to Speech technology, and it’s suitable for various embedded platform. Delta can provide various versions of speech corpus and synthesizing speed, according to system resource. More system resource, the synthesis speech quality is better; less system resource, the smaller size speech corpus and faster synthesizing speed is also workable. Besides Mandarin Text to Speech, Delta also provides Mandarin/English mixed lingual version.

Entering address into GPS device by hands is  not convenient for users, especially while driving a car. Voice input provides hands-free and eyes-free to entering destination address.

Scenario1 – 3 Steps
Says “Taipei Neihu”, then “Ruey Kuang Road”, then “186”

Scenario2 – One Shot
Says “Taipei Neihu, Ruey Kuang Road, 186”