Speckle Noise Detection and Removal for Laser Speech Measurement Systems
Round 1
Reviewer 1 Report
-The paper proposes a novel automatic impulsive noise detection and removal method in laser measured speech.
-I would recommend to have a detailed proofread of all the paper as there are some typos.
-In Figure 9, the signal shown in (e) is not described in the caption.
-In the speech used from the Librispeech ASR corpus it is not clear if the authors used speech of just one speaker or of many speakers.
-Very interesting laser speech measurement approach that could be applied in areas like security and military. Have you considered applications in speech therapy as an unobtrusive way to support patients or to enhance automatic translations?
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
This paper examines laser speech measurement, analysis of audio without need for contact, used in some security applications. The paper examines 14 characteristics of impulsive noise in laser measured speech and proposes an automatic impulsive noise detection and removal method, via decorrelation based on a linear prediction (LP) model. Located bad samples are replaced by estimated LP samples. This automatic speckle noise detection and removal method outperforms other related methods across a wide range of degraded audio signals.
The paper is professionally done, and both very well written and well organized. It was a pleasure to review. The quantative improvements are small, but exist.
The significsnce of this work may not be large, as this method of speech acquisition is subject to much noise. The quality of even imprioved speech is not high in many cases.
Specific points:
.. the noise-to-signla ratio (NSR)..->
.. the noise-to-signal ratio (NSR)..
(Note: most literature uses SNR (signal-to-noise ratio), instead)
.. laser speech measurement system demodulates a speech signal .. ->
.. laser speech measurement system obtains a speech signal ..
(The word “demodulate” means to extract a modulating signal from its carrier, which does not apply here; its use later is OK, after one explains where the modulation comes from)
..measuring the vibration caused by the sound. - well, that is also what a microphone does; you may wish to distinguish here
Fig. 1 has undefined acronyms: BS, PBS and AOM
..In the previous work, .. ->
..In previous work, ..
..used the decorrelation method based on linear prediction (LP) model ..->
..used a decorrelation method based on the linear prediction (LP) model ..
..In this article, we presented a .. ->
..In this article, we present a ..
..is fatal, so, for laser ..
..is fatal; so, for laser ..
..so it is impossible to ..
..so it is very difficult to ..
(The text here views lots of difficult tasks as impossible, which may be too strong a statement)
..methods[10] ..
..methods [10] ..
..In view of the periodic nature 121 of speech, .. - no, much of speech is aperiodic
..ACF-based algorithms have good at ..
..ACF-based algorithms are good at ..
(i.e., put a space before each [ )
..where, the integer variable ? is ..
..where the integer variable ? is ..
(No comma here; also, do not indent after an equation)
..Levinson–Durbin algorithm [26]
..Levinson–Durbin algorithm [26].
- 4 has an extra “? = ? + ? ”
.. inverse and transform the ..
.. invert and transform the ..
.. is defined as follow ..
.. is defined as follows ..
In section 3.2, eq. 11 does not account for the impulsive excitations of the vocal cord closures, whicjh would require use of a pitch detector
.. Janssen [33] suggest using ? =3???? + 2. –
- I do not see any ref. 33 (list stops at 32)
- LP order for speech is usually 10 for 8-kHz telephone speech; it can be hiogher for more wideband speech
- Why link P to window size? It usually corresponds to the resonance spectral detail desired
.. Typically, this 362 happens every 10 seconds. – this is highly variable; 10 s may be an average
Ref. 16 uses all capital letters
Name LV should be Lv (line 375)
.. detection system that constructed by ourselves for ..
.. detection system that we constructed by ourselves for ..
.. was aimed at the paper cup .. – what cup? This should be explained
.. from the Librispeech ASR corpus .. – why use this corpus, designed instead for ASR?
.. optical path. to get ..
.. optical path, to get ..
.. do not seem surprising are perfectly satisfactory ..
.. do not seem surprising and are perfectly satisfactory ..
.. for ordinary electronic microphone.
.. for ordinary electronic microphones.
Refs 4 and 6, 21-23 have no named source
Ref 14 name in italics
Author Response
Please see the attachment.
Author Response File: Author Response.docx
