Dragon voice recognition resources and news

Background

This is an evaluation we have made of the new Plantronics Savi Office WO100 Wireless headset system for use with Dragon NaturallySpeaking.

 

Plantronics have established that the market is moving quickly to Unified Voice Communications. Plantronics have developed this enterprise – grade headset system to connect to both to a user’s desk phone and PC, providing business critical sound quality necessary in today’s working environment. The Savi Office headset system allows a user to ”mix and match” calls between a PC and desk phone and even to connect a VOIP conversation with a desk phone conversation.

The Savi Office is available in 2 distinct headset styles.

Savi Office WO100                                                         Savi Office WO200

We only evaluated the Savi Office WO100. We understand that the performance of the 2 units is identical (apart from less talk time with the WO200) but when we have had chance to do further testing on the WO200, we will update this review. This has now been done. See addendum at the end of the report.

We currently offer a hybrid CS60 Dect Wireless as well as a hybrid CS70 microphone with GN Netcom Select CTI switch for this purpose. Both with or without and additional USB sound adapter.

Whilst both systems gave excellent accuracy (with the modified CS70 having the edge on noise cancellation) and enabled easy switching between the telephone and PC, one problem was that if the headset lifter was fitted, as soon as the user would remove the headset from the cradle to use speech recognition or VOIP, the headset lifter would lift the telephone’s handset from the cradle. The only way to stop this was to disable the headset lifter but this meant that calls could not be taken away from the desk.

This problem has been overcome with the Savi Office.

Another problem was that for users with poor PC sound cards, an additional USB sound adapter was necessary at additional cost.

Again, this problem has been overcome with the Savi Office as it has been designed to connect to the PC’s USB port.

Setup

For this evaluation, the Savi Office was connected to the a USB port of the PC (Vista Home Premium installed) and to a BT Converse 1100 desktop telephone exactly according to the quick start guide and interactive setup guide on the Plantronics website.

The version of the Dragon NaturallySpeaking software was Professional version 10.1

The Personocall software provided with the Savi Office on CD-ROM was not installed.

Evaluation results

We evaluated the Savi Office in terms of:-

1. Clarity of the instructions in the quick setup guide
2. Clarity on both sides of a telephone call
3. Dictation accuracy in Dragon NaturallySpeaking with no noise in the background
4. Dictation accuracy in Dragon NaturallySpeaking with simulated office noise in the background
5. Voice clarity with VOIP (Skype)
6. Wearing comfort
7. Use with the Plantronics HL 10 headset lifter

Our assessment for point 7 is somewhat subjective as there is no “method of test” for wearing comfort.

Results

Clarity of instructions in the quick setup guide

We did find the quick start guide to be a little confusing, in particular:-

5-2 this tells the user to set speaking and listening volumes to setting 3. There were no numbers on the base unit dials on the 2 units I received. Instead, these were set based on dial clicks.

5-3 Says “With your headset on, press the call control button on headset” – I didn’t realise I had to power on the unit by pressing and holding down the call control button for a few seconds until green lights flashed behind the call control button. On our unit, the headset didn’t switch itself on automatically when it was removed from the cradle.

We found the online interactive setup guide to be excellent.

Clarity on both sides of a telephone call

The handset of the BT Converse 1100 was removed in order to place a telephone call.

Good clarity on both sides of a telephone call were achieved with the settings below when wearing the WO100 both close to the base station and several metres away:-

Speaking volume	Setting 2
Listening volume	Setting 1
Configuration switch	Setting A

 
The instructions recommend setting 3 for both speaking volume and listening volume. However at settings 3 and above, there was a lot of noise feedback in the headset which was unacceptable. There was also a lot of interference in the headset which was unacceptable.

Dictation accuracy

A new user profile was created in Dragon NaturallySpeaking for the Savi Office using the above settings established for optimum clarity on the telephone.

The following are screen captures of the new user profile setup in Dragon NaturallySpeaking:-

Note the extremely low flat yellow noise floor in the above screenshot. This is a highly desirable result for a microphone for use with Dragon NaturallySpeaking.

Note the high speech to noise ratio of 29. This is a highly desirable result for a microphone for use with Dragon NaturallySpeaking.

And when the “play” button was pressed in the above screen, playback was found to be extremely clear with very high voice fidelity.

Next, a five-minute enrolment text training passage was read in order to allow Dragon to acclimatise my voice with the Savi Office. The enrolment text was completed easily with no re – reads of sentences required as is the case with some microphones.

Next, dictation accuracy was measured by reading the “Rainbow Passage” 3 times and averaging the result. See Appendix I for the procedure.

The results were as follows:-

Speech recognition mode	Accuracy
Dictation accuracy (percentage accuracy in dictating the “rainbow passage” with no background noise (average of 3 readings)	98.1%
Dictation accuracy (percentage accuracy in dictating the “rainbow passage” with simulated office noise in the background (average of 3 readings)	97.8%

 
Next, background noise cancellation was subjectively assessed as follows:-

A recording was made in Vista with the Sound Recorder

The first part of the recording was made with normal ambient noise in the background

The second half of the recording was made with the above typical office noise soundtrack introduced in the background

The recording was assessed for the audibility of the background noise

Whilst the simulated office noise in the background is clearly audible in the second half of the recording the above results show that the reduction in word recognition accuracy in going from no background noise to simulated office noise in the background is quite small. Therefore, despite the background noise being audible, the headset must “recognise” the voice much better than it “recognises” background noise.

Voice clarity with Skype VoIP

A test call was made to Skype’s Echo/Sound Test Services and played back through USB powered speakers.

Playback clarity was judged to be excellent.

Wearing comfort

The Savi Office headset came with an over the ear clip as standard. We found the headset to be easy to fit onto the ear. Fit was reasonable although the standard gel earpiece and tab felt a little on the large side to fit properly into the ear. The tab ring was removed but this didn’t seem to make much difference. The headset needed to be rotated a few times and pushed into the ear canal to achieve a good consistent fit. Otherwise, the headset felt loose and flapped away from the face. After a few attempts, we mastered the fit issue and had no problems from then on.

Once fitted, the tip of the boom was positioned close to the corner of the mouth (as required for optimum speech recognition accuracy) and the headset felt quite comfortable to wear.

We removed the gel earpiece and tab and replaced it with one of the foam ear cushions in the Savi WH100 comfort kit which was supplied in the box. However fit into the ear canal was not judged to be as good so we reverted to the gel earpiece and tab.

Next, the ear clip and replaced with the headband. Fit felt comfortable and secure. Possibly, the headband will be the best choice for speech recognition as perhaps, it will give the most consistent positioning of the tip of the microphone boom relative to the corner of the mouth during a long dictation session.

A slight negative was that the headband got in the way when the headset was placed in the base for charging as the headband rotates loosely relative to the headset itself.

Looking at the WO200 headset version on the Plantronic’s website, this looks like a user could adjust for better fit into the ear canal as it has a range of ear gel sizes.

Use with the Plantronics HL 10 headset lifter

The HL 10 Headset Lifter was placed under the handset of the BT Converse 1100 telephone.

When receiving a call, a few beeps are heard in the headset and you press the headset call control button to take the call. Pressing the headset call control button ended the call and the handset was dropped back into the telephone cradle. It was possible to seamlessly switch between the PC for speech recognition and the telephone with headset lifter.

Conclusions 

1. We feel that some of the points in the Quick Start Guide could be improved. The online setup guide is excellent but in addition, for non technical users (i.e. for most people), instructional setup real life videos (i.e. not flash animation) would be great and could provide a “competitive edge” for Plantronics.
2. Speaker and listener clarity in telephone calls is excellent although the settings that had to be used were different to those recommended in the guide.
3. The measured accuracy in Dragon NaturallySpeaking was 98.1%. This is an excellent result and is the same result that you would typically get from the best wired headset available for Dragon NaturallySpeaking.
4. The accuracy with office noise in the background only dropped to 97.9% percent. Again, this is a most excellent result and is similar in performance to a high specification active noise cancelling wired headset.
5. Voice clarity with Skype was excellent
6. Wearing comfort was acceptable although fit in the ear canal could be improved
7. The Savi Office allows a user to seamlessly switch between activating the headset lifter to take calls and the PC for VOIP or speech recognition without a problem.

Addendum

Since making the above evaluation, we have got hold of a WO200 unit. Here are the results:-

Base station configurations settings compared:-

	Savi Office WO100	Savi Office WO200
Speaking volume	Setting 2	Setting 2
Listening volume	Setting 1	Setting 0 with volume on actual headset set at maximum
Configuration switch	Setting A	Setting A

Speech recognition results compared:-

Speech recognition mode	Accuracy WO100	Accuracy WO200
Dictation accuracy (percentage accuracy in dictating the “rainbow passage” with no background noise (average of 3 readings)	98.1%	98.7%(average of 4 readings) 98.9% (average of best 3 reading results
Dictation accuracy (percentage accuracy in dictating the “rainbow passage” with simulated office noise in the background (average of 3 readings)	97.8%	98.4%
Speech to noise ratio in Dragon’s Audio Setup Wizard	29	28

Appendix I

The Rainbow Passage

The rainbow passage

When sunlight strikes raindrops in the air, they act like a prism and form a rainbow. The rainbow is a division of white light into many beautiful colours. These take the shape of a long round arch, with its path high above, and its two ends apparently beyond the horizon. There is, according to legend, a boiling pot of gold at one end. People look, but no one ever finds it. When a man looks for something beyond his reach, his friends say he is looking for a pot of gold at the end of the rainbow.

Throughout the centuries men have explained the rainbow in various ways. Some have accepted it as a miracle without physical explanation. To the Hebrews it was a token that there would be no more universal floods. The Greeks used to imagine that it was a sign from the gods to foretell war or heavy rain. The Norse men consider the rainbow as a bridge over which the gods passed from Earth to their home in the sky. Other men have tried to explain the phenomena physically. Aristotle thought that the rainbow was caused by a reflection of the sun’s rays by the rain. Since then physicists have found that it is not reflection, but refraction by the raindrops which causes the rainbow. Many complicated ideas about the rainbow have been formed. The difference in the rainbow depends considerably upon the size of the water drops, and the width of the coloured band increases as the size of the drops increases. The actual primary rainbow observed is said to be the effect of super position of a number of bows. If the red of the second bow falls upon green of the first, the result is to give a bow with an abnormally wide yellow band, since red and green lights when mixed form yellow. This is a very common type of bow, one showing mainly red and yellow, with little or no green or blue.

335 words

Count number of incorrect words = “A”

% accuracy = (335 – A)/335 X 100%