Posted on

Evolution of Speech Recognition Technology

By Sahil Chauhan for Read Write

Communication plays an essential role in our lives. Humans started with signs, symbols, and then made progress to a stage, where they began communicating with languages. Later computing and communication technologies came. Machines began communicating with humans and in some cases, with themselves also. The communication created the world of the internet, or as we technically know the Internet of Things(IoT). Here is the evolution of speech recognition technology that involves machine learning.

The Evolution of Speech Recognition Technology and Machine Learning

The internet gave rise to new ways of using data. Using this, we can communicate directly or indirectly with machines by training them, which is known as Machine Learning. Before this, we have to access a computer to communicate with machines.

Research and development are beginning to eliminate some of the use of computers to a great extent. We know this technology as Automatic Speech Recognition. Based on Natural Language Processing (NLP), it allows us to interact with machines using our natural language in which we speak.

The initial research in the field of Speech Recognition has been successful. Since then, speech scientists and engineers aim to optimize the speech recognition engines correctly. The ultimate goal is to optimize the machine’s interaction according to the situations so that error rates can be reduced and efficiency can be increased.

Automatic Speech Recognition and its Applications

Automatic Speech Recognition(ASR) technology is a combination of two different branches – Computer Science and Linguistics. Computer Science to design algorithms and to program and Linguistics to create a dictionary of words, sentences, and phrases.

Generating Speech Transcriptions

The first stage of development starts with speech transcriptions, where the audio is converted into text, i.e., speech to text conversion. After this, the system removes unwanted signals or noise by filtering. We have different voice speeds while saying a word or sentence, so the general model of speech recognition is designed to account for those rate changes.

Later the signals are further divided to identify phonemes. Phonemes are the letters that have the same level of airflow, like ‘b’ and ‘p.’ After this, the program tries to match the exact word by making a comparison with words and sentences that are stored in the linguistics dictionary. Then, the speech recognition algorithm uses statistical and mathematical modeling to determine the exact word.

Speech Recognition systems are of two types, at present.

One type of system is accomplished with learning mode and other as a human dependent system. With developments in Artificial Intelligence(AI) and Big Data, speech recognition technology achieved the next level. A specific neural architecture called long short – term memory bought a significant improvement in this field. Globally, organizations are leveraging the power of speech at their premises at different levels for a wide variety of tasks.

Speech to text software can be used for converting audio files to text files.

Speech to text software includes timestamps and confidence score for each word. Many countries do not have their language embedded keyboards, and a majority of people do not have an idea of using a specific language keyboard, though they are verbally good at it. In such cases, speech transcriptionshelp them to convert speech into text in any language.

Real-time Captioning System — Captions on the go.

The other use of this technology is in real-time. Tech done in real-time is known as Computer Assisted Real-Time translation. It is basically a speech to text system which operates on a real-time basis. Organizations all over the world perform meetings and conferences.

For maximum participation by global audiences, they leverage the power of live captioning systems. The real-time captioning system converts the speech to text and displays it on the output screen. It translates the speech in one language to the text of other languages and also helps in making notes of a presentation or a speech. These systems convert speech to text that is also understood by hearing-impaired people.

Voice Biometric System — A Smart way to Authenticate

Apart from speech to text, the technology spreads its branch into the biometric system, which created voice biometrics for authentication of users. Voice biometric systems analyze the voice of the speaker, which depends on factors like modulation, pronunciations, and other elements.  

In these systems, the sample voice of the speaker is analyzed and stored as a template. Whenever the user speaks the phrase or sentence, the voice biometrics system compares them with the stored template and provides authentication. However, these systems are facing a lot of challenges. Our voice is always affected by physical factors or emotional state.

The recent developments in biometric voice systems operate by matching the phrase with the sample. After this, it analyzes the voice patterns by taking psychological and behavioral voice signal into consideration. Also, the developments in voice biometrics technology are going to help enterprises where data security is a significant concern.

Using Speech for Analytics

Analytics play an essential role in the development of speech recognition technology. Big data analysis created a need for storing voice data. Call centers started using the recorded calls for training their employees. Since customer satisfaction is now the primary focus of organizations around the globe. Now, organizations want to track and analyze the conversation between executives and customers.

With Call Analytics applications, organizations can monitor and measure the performance and analytics of call. This call analytical solution enhances the performance of services provided by call centers. Through this, one can classify their customers and can serve them better by giving faster and favorable responses.

Way Ahead For Speech Recognition Technology

Research in speech recognition technology has a long way to go. Until now, the program can act on instructions only. Human communication feel does not exist entirely with machines. Researchers are trying to inculcate the human responsiveness into machines. They have a long way to go in the innovation of speech recognition technology.

The primary feature of research concentrates on how to make speech recognition technology more accurate. For human language understanding, we need more accuracy. For example, a person raised a question, “how do I change camera light settings?” This question technically means that the individual wants to adjust the camera flash. So significant concentration is on understanding the free form language of humans before answering specific questions.

So overall, machine learning with speech recognition technology has already made its way into the organizations globally and started providing effective and efficient results. Very soon we might be seeing a day where the automated stenographer would get promoted and start taking an active part in organizing the meetings and presentations.

Posted on

Thinking about getting speech recognition for your small practice?

There are some pretty impressive stats about the amount of time physicians spend documenting.  According to a study on PubMed.gov, 43% of doctors spend their workday on data entry and click an average of 4,000 times per day during documentation. Working as a physician should not be a desk job, but these statistics demonstrate how it can be. As for now, there isn’t a way to avoid the 7.2 million words a doctor will document within the year. However, there is a way to lessen the burden: speech recognition.

If you’ve been researching speech recognition software for your small physician practice you’ve probably heard about the top solutions from M*Modal and Nuance. M*Modal’s Fluency Direct is a previous Best in KLAS recipient for Speech Recognition: Front End EMR. Still, Nuance’s Dragon Medical Practice Edition is one of the most popular solutions and achieves 99% accuracy right out of the box without voice training. Dragon Medical One is the newest speech recognition option from Nuance. Dragon Medical One is cloud-based and can be used on a wide range of Windows® devices.

At a glance, each speech recognition solution has great pros. Who wouldn’t want award-winning or highly accurate speech recognition. So, we dove a little deeper and compared the three popular options in small practice speech recognition. Let’s start with Fluency:

M*Modal’s Fluency Direct for Practice

Key Features:

  • Create, edit and sign clinical notes directly within EHR templates
  • Natural Language Understanding technology for contextual understanding of the physician narrative to improve accuracy
  • Designed for practices with 10 users or less
  • Covers sub-specialty medical terminology and supports all regional accents
  • Includes a microphone
  • Easy, click-once installation from a web address and minimal software training requirements
  • Auto-updates and upgrades
  • Single cloud-hosted voice profile
  • Customized macros can be used to enter often-dictated text
  • Minimal training
  • Machine learning based on the collective voice profiles of over 200,000 clinicians
  • Patient information is immediately available in the EHR
  • Adjusts on-the-fly to differences in cadence, accent, dialect and medical terminology
  • Edited through voice commands followed by finalization and electronic signature
  • HIPAA Compliant

Online review:

“Fluency Direct has been an awesome addition to my work flow. As a primary care physician, dictation, transcription, or voice recognition has always seemed like something out of reach. Having embedded M*Modal in my work flow has transformed the way I approach my patients, process my documentation, and rapidly and efficiently close my encounters. I cannot speak any more quickly in any noisier environment than some of my work spaces, speaking quietly yet accurately having my thoughts transcribed into my documentation. I spend much more time face-to-face with my patients using only cryptic notes to key me into a more complete documentation of my encounter once I leave the room. With my microphone as my navigator, I can move about my chart using voice commands to move to different sections, voice activated macros to fill common dialogue, and have been able to “pre-round” on my patients saving vital information to easily post after intake is complete. I do not type well, but you cannot type this fast!”


Management, Pediatric Medicine
Clinic size: 9

Dragon Medical Practice Edition 4

Key Features

  • Achieves 99% accuracy out of the box without voice profile training
  • Available for independent practices of 24 or fewer physicians
  • Combines 90 medical specialty and sub-specialty vocabularies with acoustic models based on audio, syntax, style and structure
  • Regional accent support
    • Advanced adaptation techniques and accent-specific acoustic models
  • Advanced Deep Learning technology constantly learns and adapts to voice and environmental variations – even during dictation – to refine performance
  • Dictate for real-time speech to text or transcribe audio recordings
  • Movable DragonBar provides easy access to popular features and collapses when not in use
  • Hybrid touch and keyboard interaction work for controlling the DragonBar
  • Customized macros for frequently dictated text can be created with a voice command
  • Compatible with Windows 10 touchscreen devices
  • Dictate within applications and EHR textbox fields, or use the Dragon dictation box to compose content
    • insert auto-texts
    • navigate template fields
    • dictate and edit – and transfer text with a simple voice command
  • supports HIPAA requirements
  • easy access to popular help searches and topics
  • Includes a full library of AutoTexts for standard notes and “medical normals” by body system
  • Automatically detects hardware resources and determines the best use of infrastructure
  • No internet connection required. Locally-installed speech recognition ensures uninterrupted access
  • Automatically detects poor audio input and alerts the user with advice to remedy the situation and ensure high-accuracy results.

Reviews

“With Dragon Medical Practice Edition and the EHR, we can generate the code that truly reflects the work we do for complex patients. Our Level 4 encounters increased from 3% to 14% over a six-month period.”

Chuck Stillwaggon
Practice Administrator, Orthopedics Northwest, PLLC, Yakima, WA

“I use the EHR for a lot of point-and-click, but for the subjective information, as well as my conclusions and impressions, that’s where Dragon Medical Practice Edition shines…I would never go back to life pre-Dragon Medical.”

Andrew Fireman
MD, Cardiologist, AMS Cardiology, Abington, Pennsylvania

Dragon Medical One

Key Features

  • provides secure, accurate, and portable, cloud-based clinical speech recognition across a wide range of Windows® devices
  • Nuance® Healthcare ID puts power in the clinician’s hand with the ability to personalize their experience and gain access to new features, products, and services as they emerge
  • access their personal voice profile in a growing catalog of mobile productivity apps. Including:
    • secure communication
    • care coordination
    • clinical reference tele-medicine
    • population health
  • comes with a secure online analytics portal to track clinician efficiency, productivity and workflows
    • Use it to help determine what is working and pinpoint areas of improvement for informed decision making
  • no per-device limits so clinicians can stay productive anywhere
  • complete their patient notes at any available workstation, with or without a hard-wired microphone, as soon as they meet with each patient
  • PowerMic Mobile is a smartphone compatible app available via the App Store and Google Play Store to simplify the documentation process

Check out these case studies by Nuance:

The difference between Dragon Medical Practice Edition vs. Dragon Medical One (DMO)

If you’re torn between two great speech recognition solutions from Nuance and want to know the difference, we’ll make it simple for you: the payments and device access. Practice Edition is installed on one local device and DMO is stored in the cloud. This means DMO can run on any computer that has access to your medical voice recognition profile.

DMO is also paid through a monthly subscription of $99.00. The monthly payments enable automatic backup of your profile and access to new update that enhance accuracy and language modeling.