LumenVox

LumenVox specializes in the development of a speech automation and authentication solutions for self-service and customer-agent interactions. LumenVox's product stack includes call progress analysis, voice biometrics, speech recognition, and text-to-speech solutions. LumenVox offers support for numerous platforms, including Asterisk, Avaya, Cisco, Dialogic, Enghouse, and Genesys. The LumenVox solution has various applications, such as fraud detection, proof of life, and password resetting.

LumenVox solutions overview (marketing material)

LumenVox's call progress analysis (CPA) solution is differentiated from answering machine detection (AMD) technology that uses energy-based audio detection to determine information. Instead, the company's proprietary CPA speech recognition technology uses voice activity detection (VAD) software based on machine learning. The speech recognition engine detects the tone of the answering machine and indicates to the user to leave a message, in order to ensure that the message is delivered in its entirety.

LumenVox's active authentication aims to reduce hold and agent wait times by assigning authentication tasks to an automated process early in the customer-agent interaction. The solution also protects applications with biometrics-based identity verification, preventing unauthorized access to information and accounts. Moreover, according to the company biometric authentication reduces costs associated with operations and fraud. LumenVox's platform enables the setup and maintenance of diverse enrollment and authentication workflows via web Based APIs (including SOAP and RESTful interfaces) that allow for direct connections to existing applications. The LumenVox voice biometric engine uses the following process to actively authenticate a person:

1. A person's claimed identity is verified via an organization’s predetermined methods
2. The person records their voiceprint as prompted by an organization’s IVR or mobile application
3. When the person is requested by an organization to authenticate, they are asked to repeat a random set of words or a passphrase to provide the voice biometric engine with enough information to authenticate them
4. The person's voice is then compared to the recorded voiceprint and verified

LumenVox's passive voice biometrics is a customer service solution that authenticates the user as they interact with an organization's IVR, live agent, or self-service channel. Initially the user is required to register about thirty seconds of speech within a call center environment, and thereafter a minimum of eight seconds is required for authentication purposes. The process involves the following steps:

1. Customers interact with the IVR and/or agent
2. During the call, the caller's speech is analyzed and a voiceprint is created
3. IVR – callers who interact with a speech-enabled IVR can be authenticated in the IVR. CTI ensures that the authentication level (score) is passed along to the agent

If the caller is not authenticated in the IVR, their first words are analyzed and a voiceprint is created. In the background, the newly created voiceprint is compared with the voiceprint on file. In the event of a match, the caller is authenticated and the agent is notified with a green light. In the event of a mismatch, the agent is notified that the authentication has failed with a red light. If a voiceprint is not present in the system, the agent is advised to authenticate the caller manually. Once the caller is authenticated, the voiceprint is stored in the system to be used for future calls.

LumenVox's automatic speech recognition (ASR) engine is built on an end-to-end DNN architecture. This engine accelerates the process of adding new languages and dialects and provides a toolset to expand the language model for the purpose of serving a more diverse user base. LumenVox's adoption of standardized protocols and development tools make it easier to replace existing ASR systems within infrastructures with the solution. The company's media server supports MRCP (media resource control protocol) v1 and 2.

The LumenVox TTS Server can transform text into speech, allowing computers or IVR applications to pass information to a user in this format, which can be used in speech telephony platforms or software applications. Common uses of this technology include telephone systems, automotive applications, and programs to assist the disabled. It can be especially useful for reading dynamic data that cannot be pre-recorded, such as reading live text from the web or a street address from a database. The LumenVox Text-to-Speech Server can be run on 32- or 64-bit versions Windows or Linux (Red Hat or CentOS), on any machine with a processor of 2 GHz or faster, with at least 4 CPU cores and 2 to 8 GB of RAM. It is designed to be controlled via LumenVox's C/C++ API or MRCP versions 1 or 2. LumenVox's text-to-speech software includes male and female voices in the following language variants:

• American English

• Australian English

• British English

• Canadian French

• Castilian Spanish

• Danish

• Dutch

• European French

• European Portuguese

• Icelandic

• Indian English
• Italian
• Latin American or North American Spanish
• Norwegian
• Polish
• Romanian
• Russian
• Standard German
• Swedish
• Turkish
• Welsh
• Welsh English

LumenVox's clients include a US national domestic airline, a global cloud communications company, a multinational mobile telecommunications company, an NA-based provider of home health care services and products, and more.