Voice Recognition For Security Systems

Voice and speech recognition have progressed leaps and bounds in recent times. Humans depend too much on mobiles these days. According to a study, most people on average spend 3 hours and 15 minutes on their phones. That’s a lot of human dependence on a single piece of tech. Even multinational conglomerates across the globe are coming to realizing that a smooth and efficient human to computer interaction is the need of the hour. They have identified voice recognition software as a much-needed tool to streamline the tasks that are otherwise conventionally done. 

As people become increasingly comfortable with biometrics, voice authentication is finding wider application across industries, including healthcare, banking, and education. The voice biometrics market is set to grow at an explosive CAGR of 19.4% between 2017 and 2021. Voice recognition systems monitor the cadence and accent, as well as indicate the shape and size of the larynx, nasal passages, and vocal tract of a person, to help identify and authenticate the individual. 

The Million-Dollar Question: How Does Voice Recognition Work? 

Voice Recognition means making a computer understand human speech. It is done by converting human voice into text by using a microphone and a speech recognition software. The basic recognition of speech system is shown below:

Sampling of Speech Signal

1. Speech to text conversion 

When sound waves are fed into the computer, they need to be sampled first. Sampling refers to breaking down of the continuous voice signals into discrete, smaller samples- as small as a thousandth of a second. These smaller samples can be fed directly to a Recurrent Neural Network (RNN) which forms the engine of a speech recognition model. But to get better and accurate results, pre-processing of sampled signals is done.

 

 Sampling of Speech Signal

2. Pre-processing of speech 

Pre-processing is important as it decides the efficiency and performance of the speech recognition model. Sampled waves are usually as small as 1/16000th of a second. They are then pre-processed, which is breaking them into a group of data. Generally grouping of the sound wave is done within interval of time mostly for 20-25 milliseconds. This whole process helps us convert sound waves into numbers (bits) that can be easily identified by a computer system. 

3. Recurrent Neural Network (RNN) 

Inspired by the functioning of human brain, scientists developed a bunch of algorithms that are capable of taking a huge set of data, and processing that it by drawing out patterns from it to give output. These are called Neural networks as they try to replicate how the neurons in a human brain operate. They learn by example. Neural Networks have proved to be extremely efficient by applying deep learning to recognize patterns in images, texts and speech. 

Recurrent Neural networks (RNN) are the ones with memory that is capable of influencing the future outcomes. So RNN reads each letter with the likelihood of predicting the next letter as well. For example, if a user says HEL, it is highly likely that he will say LO after that, not some gibberish such as XYZ. RNN saves the previous predictions in its memory to accurately make the future predictions of the spoken words. 

Using RNN over traditional neural networks in preferred because the traditional neural networks work by assuming that there is no dependence of input on the output. They do no use the memory of words used before to predict the upcoming word or portion of that word in a spoken sentence. So RNN not only enhances the efficiency of speech recognition model but also gives better results.

Speech recognition model using RNN

4. RNN Algorithm 

 

 Steps Involved in RNN Algorithm

a. The input states: 

Xt = input at time t 

Xt-1 = past input 

Xt+1 = future input 

b. St 🡪 hidden state. It is the hidden memory. It stores the data of what things took place in all the previous or past time steps. It is calculated as: 

St = f(U*Xt + W*Xt-1) 

c. The output states:

Ot🡪 output at the step t. It is calculated exclusively based on the memory at time ‘t’. It is calculated as: 

Ot = softmax(V*St) 

As shown in the figure, the RNN uses (U,V,W) parameters. It implies that by passing various inputs at different steps, the same task is being done at every step. This limits the number of parameters to be learned. Even though there is an output at each time step, dependence on the task to perform is not required. 

To make it easier to understand, consider an example where we have to predict the output of a sentence. To do so, we won't concern ourselves with the output of each word, but with the final output. Same implies for the inputs as well, that is, we do not need input at each time step. 

5. Training An RNN 

So far, we know that in RNN, the output at a certain time step not only depends on the current time step but also on the gradients calculated in the past steps. Consider an example where you have to calculate the gradient at t=6. In order to do so, you will have to back propagate 5 steps and sum up all the gradients. This is called Back propagation through Time (BPTT) and we employ this algorithm to train an RNN. 

This method of training an RNN has one major drawback. This makes the network to depend on steps which are quite apart from each other. This problem of long term dependency is sorted by using other RNNs like LSTM. 

6. LSTM 

We know that RNN cannot process very long sequences. To overcome this problem, scientists came up with Long short- term memory or LSTM. While RNN have only one structure, LSTM have four. They consist of cell state that allow any information to flow through it. By applying gates, any information can also be added or removed. 

LSTM employ three types of gates: input gate, output gate and forget gate. These three gates together protect or control the cell state. LSTM also uses sigmoid functions that only give two outputs. Either they will pass every information that was given at the input or no input information will be passed at all. 

This is where LSTM is better than RNN as using cell states we can control the long- term dependencies. 

 

 LSTM Model for Voice Recognition Software

Types of Speech Recognition Software :

Speaker-dependent voice recognition software: They are heavily dependent on the speaker as they need to learn and analyze the characteristics of the user’s voice. Once provided with enough data to recognize the voice and speech patterns, they can be used as highly efficient dictation software. 

Speaker-independent voice recognition software: They do not depend much on the speaker’s voice pattern, as they are trained to recognize anyone’s voice. Naturally, they are not as efficient as the speaker-dependent software and hence are more commonly found in telephone applications. 

Command and control voice recognition software: These systems are used to navigate and control devices using voice commands. Tasks such as starting the programs, browsing through websites and other functions can be easily accomplished. 

Discrete input voice recognition software: They aim for high accuracy of word identification. They do so by requiring a pause after each word is spoken. This limits their efficacy to around 60-80 words per minute. 

Continuous input voice recognition software: They are designed to analyze a stream of continuous word flow. As compared to other software, they consider fewer words and hence find their application mostly in medicine and healthcare. 

Natural speech input voice recognition software: They are capable of understanding words that are spoken fluently and can understand as high as 160 words per minute. 


Advantages of voice biometric authentication: 


Satisfactory customer experience: Common user authentication methods such as passwords, PINs, questionnaire-based on personal information, etc. can be a setback in customer experience as it makes the customer wait for some time before they are able to avail of any services for which they contacted the call center. Voice authentication, on the other side, occurs passively in the background, enabling customer’s experience to be the priority, while recognizing speech without interrupting the conversation.

Increased security: Layers of security arises through voice authentication measures. The method of authentication acknowledges not only one’s voice but also their character and other factors that affect the sound. This reduces fraud as the biometric channel can quickly identify a different voice.

Reduces operational costs: Voice authentication reduces the operational cost of call centers and even banks. It saves them millions by removing many processes that are involved in ancient authentication techniques. On an end to end conversation, it can recognize the customer’s voice to verify the identity without the requirement of standard questions.

Protects brand reputation: Companies that use voice authentication have the advantage of protecting their brand reputation. In this era and age, where consumers are adopting online services, their trust builds when organizations ensure encryption of their account data and their privacy. Consumers can quickly shift towards competitor’s services if their information is leaked, and their confidence breached.

Accuracy: Voice authentication is more accurate and reliable than the use of passwords that can be easily lost, changed, or guessed. It is like fingerprints, which nobody can have someone else’s. In other words, you cannot forget or recreate a voice, as in the case of passwords. Although the sound might be affected by several factors, it is much more reliable and convenient. 


Addressing the Challenges of Voice-Based Authentication:

While the voice of an individual is unique, secure authentication through voice recognition can be a challenge in some cases – for instance, if the user has a sore throat or cold. It is therefore important to prevent unauthorized users from hacking into the database by mimicking someone else’s voice.
The ideal way to do this: whitelist the voiceprints and store them in the Active Directory – a process wherein a customer who uses the voice recognition system is enrolled into a whitelisted member database, and his or her voice print is used as valid voice print for authentication. 
Using the Active Directory, the unique voice cadence of each enrolled member is compared to both a whitelist of valid customer voiceprints and a blacklist of known fraudster voiceprints. While whitelist authentication is underway, passive fraud detection can be equipped to return an alert in real time - if the caller’s voice is a match to a record in the blacklist database.

A security expert Glen Greer notes, voice recognition for access systems is not terribly reliable. "It is my experience that voice-based systems have very high error rates, particularly in terms of false rejections."

Many challenges affect its accuracy. These include poor-quality voice samples; the variability in a speaker's voice due to illness, mood, changes over time; background noise as the caller interacts with the system; and changes in the call's technology (digital vs. analogue, upgrades to circuits and microphones, etc.). 

Another critical issue is the lack of established international standards. "We need a standard application programming interface (API) to reduce the issues with cost, interoperability, time-to-deployment, vendor lock-in, and other aspects of building applications," observes Markowitz. "So far, law enforcement, telecommunications services, and financial services in different countries, are using speaker verification but an API standard will make it easier and more attractive for integrators and others." 

How Enterprises are Using Voice Biometrics for Authentication? 

Voice authentication offers a flexible and cost-effective form of biometric authentication as it does not require hardware integration that might be needed in the case of other modalities such as fingerprint matching or retinal scan. Enterprises can leverage voice recognition to enhance security and provide a choice to customers in terms of how they wish to authenticate themselves. Some of the use cases of voice-based authentication across industries include:

Entertainment: Voice recognition can be used to change TV or radio channels, open and close screens, and play movies. It can also help personalize customer experience. For instance, services such as Netflix and Hulu can be personalized by determining the age of the user through voice analysis, enabling them to access age-appropriate content.

Healthcare: The global healthcare biometric market is expected to reach USD 14.5 billion by 2025, according to a recent report by Grand View Research, Inc. In an industry where data security is paramount, physicians can use voice biometrics to dictate and record patient’s health conditions directly into the system and securely retrieve patient’s personal history. This can significantly benefit patients who need to share medical records between various doctors. The system can also help dramatically reduce fraud for providers and payers by automating payment collection, and improve patient satisfaction by offering an additional payment option.

Banking: Customers can use voice authentication to operate bank lockers. Banks, on the other hand, can leverage the system to enable highly secure and advanced voice-based payments. With fraud on the rise, credit card companies and banks such as Citibank and ANZ use voice biometrics to proactively identify fraudsters and authenticate callers at their call centers. 

Education: Educational institutions can use voice recognition to provide flexibility to students with visual disability, helping them take online exams using voice authentication. 
Independent Software Vendors (ISV): For ISVs, voice authentication can enable enterprise sign on mechanisms such as those based on Active Directory, enabling authentication uniformity across enterprise applications and strengthening compliance with accessibility standards. 

Where is the Future of Authentication heading? 

As the technology goes main stream, the advantages of voice recognition are becoming clearer. It paves the way for greater efficiencies and stronger security. Paying bills through voice recognition, for instance, speeds up the process, and eliminates manual entry of password and other details for improved accuracy and consumer satisfaction. Consumers’ increasing dependence on voice search and organizational platforms with machine-to-machine communication capabilities are set to considerably impact the future of commerce, payments, and home devices. This presents enterprises with an opportunity to leverage the trend to their advantage. Once a customer is enrolled in the voice recognition system, his/her voiceprint can be seamlessly accessed across a company’s support channels, resulting in a seamless customer experience.