Voice recognition has become the preferred technology for remote authentication because of the advances made in telecommunications and networking and its ease of integration into existing systems. Unlike speech recognition, voice recognition is a dynamic process and lasts for several seconds at a time.
The system uses audio capture devices to record voice data at the time of enrollment and then extracts unique features from this voice data to create a template. This template is used at the time of identification or verification when an utterance is compared against the recorded template to deny or provide access to location or data.
How does a voice recognition system work?
The pioneering work in this field was done by the Swedish professor Gunnar Fant, who published a model that took into account the physiological traits of voice production. In order to gather inputs for this model Fant X-Rayed the voice tract during voice production. The information gleaned from this research was used by Dr. Joseph Perkell who expanded this theory by including the role of the tongue and the jaw during voice production.
The effectiveness of voice recognition today comes out of decades of research by hundreds of scientists and engineers working on statistics, linguistics, semantics, predictive algorithms and audio processing.The voice tract consists of the airway and the soft tissues cavities that play a major role in how sound is emanated. This in conjunction with the movement of the larynx, tongue and the jaw provides a unique tonality to each individual’s voice which is leverage by the existing biometric systems to identify an individual.
Text dependent and text independent recognition
Voice recognition factors text independent and text dependent speech while designing the voice recognition system. In the text dependent model the individual has to utter a password or a sequence of numbers to pass muster with the security systems. The text independent model on the other is free form and more dynamic in nature. This information is recorded and then analyzed in wave form, with time on the horizontal axis and loudness on the vertical axis. In particular characteristics like the quality, duration, pitch and intensity of the signal is studied for identification purposes.
Text Dependent and Text Independent Recognition
Applications of voice recognition
The voice recognition system can be used productively in the following cases –
Playing back information
The voice based biometric system can be used to make the dissemination of information faster by doing away with the services of a live operator. This type of service can be used to get quick information about flight departures and arrivals in the aviation industry. This type of technology can be used to cut waiting times between calls leading to better customer service. Dublin Airport was able to service 30 percent more customers with the use of this technology without any corresponding increase in staff. Dublin airport used this technology to route calls seeking information about departures and arrivals to an automated system.
Handling spike in call volumes
The voice recognition software can be used to handle unusual spike in call volumes. For example the call volume in the betting industry is characterized by spikes and troughs. The period before the race start is characterized by unusual call volume activity. The system can be configured in manner that calls related to placing a bet can be routed to an automated system and more complex calls routed to live operator.
Routing calls
The voice based recognition system can be used to route calls to the correct department. Standard life Insurance has successfully implemented this technology to route customers to the appropriate staff. The system helped Standard Life Insurance to cut misdirected calls by 66 % and increased the call handling capacity by over 25%. The system also revealed valuable information on why the customer was calling, helping in designing smarter systems.
Automated identification
The voice based recognition system can be implemented in the BFSI sector to provide flawless customer service. The current pin and password based security system fails to address the requirements of the modern era with its demands on security. The voice based recognition system bypasses the pin and password based security system, providing ease of access and security. Australia’s eight largest insurer ahm Health Management has successfully incorporated voice recognition biometrics to allow its customers to speak to its customer service representatives.
Interactive voice response (IVR) technology
Interactive Voice Response (IVR) SystemA normal IVR system is based on a four or five digit pin number. The user has to speak the pin number to get access to information or service. However, this system is cumbersome at the best of times as the customer has to change the pin number at regular intervals in order to prevent data theft.
Voice recognition or speaker recognition, uses a pre-recorded voice sample to identify whether a person is who they claim to be. A biometric based IVR is quicker and more efficient compared to a normal IVR which is based on a pin number. In this system the user is prompted to speak a random phrase, which is then analyzed and compared with a pre-recorded sample. Once the customer has cleared the authentication procedures he can be guided through the proper channels without wasting their time.
The IVR based on voice recognition has several advantages. The customer is spared from answering a number of questions as all the required information is pulled from the customer resource management once the authentication protocol is cleared. The process also quickens the process; therefore the customer can get quicker service, freeing up the IVR for the next customer.
Many customers struggle to remember their pin-numbers. Moreover, they feel frustrated at having to change the pin number at regular intervals. But with the voice recognition system the customer does not have to remember the pin-numbers as authentication is now possible by uttering a phrase. Voice recognition technology has helped many organizations improve their service and their reach by reduce cost and increasing productivity.
Advantages of a voice based biometric system
The basic advantage of voice based system is that it is cheaper than other biometric system. It can be integrated in a telephony based service to provide access to financial data, once the user is able to pass muster with the voice base recognition system.
Voice based verification system can be used to activate or reset passwords, accessing financial data, conducting telephonic inquiries and so on. According to Dr. Markowitz, voice recognition system has been successfully used in Scobey, Montana to check instances of infiltration at the border. The temperatures at the border often falls below zero making it very difficult for conventional biometric based security systems to work properly, making voice based system very valuable.
Dr. Markowitz also highlights how voice based verification system reduced thefts in the city of Baltimore, Maryland. Worried by the decreasing law and order situation, the authorities decided to equip some of the city’s buildings with voiced based security systems to check employee access after work hours. Thanks to this system there was a reduction in the number of thefts significantly improving law and order situations.
Conclusion
According to a recent Unisys-sponsored report, voice based systems have found favor with the masses because of its noninvasive nature. The report showed that voice based had the greatest acceptance followed by fingerprinting and hand geometry based biometric systems. Voice authentication systems wins hands down in poor environmental conditions characterized by presence of dust and grime. Remote verification is also possible as tracts of voice data can be verified over the phone to deny or give access to the users to systems data.
