Real Time Voice Capture Demonstration

Real Time Voice Capture Demonstration / Instantaneous Stress Value Display Download Free Demo Version

General

AVSA (Automated Voice Stress Analysis) is the result of work that began in 1995. The early years were spent in defining the mathematical treatment used to highlight graphs of stress in reference to those obtained from the polygraph, which is a stress detector used by the police in the USA. The analysis of voice stress has the advantage of not using sensors connected to the body of the individual, and works directly from sound recordings.
A knowledge base of several hundred interrogations was built years after years. Over the years, a software that displays curves of stress has been used and VSA experts have trained hundreds of specialists from 20 countries.

In recent years, studies to make the automatic evaluation of stress have been undertaken, and AVSA is the culmination of these work.

Instant measurement of stress is not significant in itself, and must be consolidated by a series of measures indicating a trend.

AVSA is not a lie detector, but it may be associated with an interrogation (or protocol) structured consisting of neutral questions to measure the basic stress, stressful questions and control questions to measure reactions of the person in this situation. Simple mathematical formulas are used to treat the responses and give an indication of the good faith of the person subjected to such interrogation.

Who is affected by the evaluation of stress?
All activities concerned with the problems of fraud (insurance, government agencies, call center, ...)
Human Relations Division
Recruitment
Sectors Occupational Medicine
individual

AVSA Software Specifications

-Uses a knowledge database provided by human experts
- Automatic computing of stress values using a neural network
- Real time stress value display (0 % to 100 %)
- Questionnaires to compute the trend of stress according irrelevant and relevant questions
- Formulas to estimate fraud level (Not deceptive, deceptive, not defined)
- Processing of recorded voice files. Display of sound and stress graphics
- Play back of sound frames or part of a frame.
- Markers of answers associated to questions. Annotations used for reports
- Prints Reports : stress graphics, stress value, annotations, and statistics
- Multi-lingual : English, french, Spanish (can be adapted to other languages that use latin characters)
- OS platforms : Windows XP,Vista,Windows 7

The Instantaneous measurement (Real Time) AVSA Free

This module aims to introduce the measure of stress from words and sentences directly captured by a microphone.
The image below shows the tab used by the instant measure of stress: The top graph is the representation of sound picked up by a microphone. Markers used to determine the quality of spoken words. Sounds too short or too weak are eliminated. Each valid word extracted is immediately subjected to treatment: the graph of stress is shown below, with color corresponding to the level calculated (range of colors ranging from green to red). The calculated value of stress is displayed to the left of this graph. A vertical band in the middle panel, displays the last 16 values calculated. A chart on a gray background displays the last 120 received values. The software can handle up to 15 words per second.
The left side of the tab is used to automatically name saved files. It starts recording of the sound sequence when you press a key on the keyboard. The color of the graph turns to yellow. It will return to green when it no longer records. The file name shall be set up by a prefix, date and time, optional, a sequence number incremented automatically.


(This image is not contractual)

Analysis of sound recordings (AVSA Pro) - Off-Line

This module analyzes sound files recorded previously, operates extraction of words or phonemes, and assigns a level of stress for each of them
"View" tab allows the processing of a sound recording in its entirety. The lower part shows the recorded sound, and the upper part shows the stress associated graphics. "Stress engine, calculates and displays the value of stress (yellow background) Each word extracted is marked with a serial number, (red background). The user can re-listen to each frame of sound present on the screen.
have the opportunity to point to the interesting answers, put an annotation, which will also serve to identify the response for the implementation of questionnaires.
Scroller allows horizontal movement along the sound file up to 5 minutes and 30 seconds of conversation, which represents in practice about 1400 words.
software deals only with files in standard WAV format, 11 khz or 22khz, 8 or 16 bits, mono. Any other format must be converted before processing.


(This image is not contractual)

Access to annotations is done by simply clicking a word in the graph. The annotations are also used when printing a report. You can choose to print only the words having an annotation and not all the words present in the conversation.

Any file scanned, annotated or not, can be printed and we get a report of one or more pages having the appearance shown on the image below. A histogram is printed at the end of each printed file, which summarizes all values of the selected sounds. In case of intensive use of software, multiple sound files can be treated in a batch process.


(This image is not contractual)

Questionnaires (AVSA Pro)

Questionnaires or protocols, complete calculations obtained during the analysis of sound recordings
A questionnaire is a set of questions in a strict order, with answers that can be only YES or NO. Some questions are used to measure stress basic (neutral). Other questions become more 'aggressive' and are to bring the accused person on the defensive.
questionnaires are of course dependent on the type of fraud or application that wants to be treated. They contain between 14 and 25 questions whose order is well defined.
A well-structured questionnaire gives a fairly good indication of the level of fraud.
historical cases treated for years, overlapping with actual convictions, lets say that the tool is effective over 90% (criminal cases).

Batch (AVSA Pro)

The software can also act in a non-visual environment, when it is launched from the "console" command prompt. The user calls the AVSA program, with parameters that indicate the filename to process, or a batch file that contains the names of several sound files. Printed reports can be made, and files containing the characteristics of each treatments are generated for subsequent use by other applications. In the "console", no human intervention is necessary.

Rates

For any information, call Alpes Software or send an e-mail
Tel: (33+) 04 76 56 07 49
Email: email@alpes-software.com
- AVSA Free version: can be downloaded from our site
- AVSA PRO "Analysis of sound recordings (including instantaneous measurement): Contact us
to acquire multiple licenses, please contact us.

Training / Specific development

Although software is easy to use, Alpes Software can provide customized training for an European company that would like. The training is to provide the foundation for the VSA, the interpretation of curves, and the use of questionnaires. For other countries, training is provided by experts VSA society Partner TDT (Truth and Deception Technologies), based in Florida.
Rate in France: 1000 Euros / day for 3 persons + travel expenses.
for specific developments, a quotation will be conducted from a specification stating the additional treatment.

Computer Configuration

The software runs on any PC running Windows XP or Vista
Processor 1.5 GHz or more
RAM Recommended: 2 Giga bytes or more
Soundcard
A good quality microphone, optional
USB interface for physical protection key.
The software is multi-lingual, English, French, and Spanish
Batch possible