Optimization of a Blind Speech Watermarking Technique against Amplitude Scaling

Table 1

Benchmark for speech watermarking.

Attack type	Attack name	Description	Parameter(s)	Default value(s)
Additive Noise	AddBrumm	It adds buzz or low frequency sinus tone to the watermarked signal to simulate the impact of a power supply		2500 : 55 to 3000 : 75	A
	AddDynNoise	It adds a dynamic white noise to the watermarked signal		20 to 40	B
	AddFFTNoise	It adds white noise to the watermarked signal in the frequency domain		256 : 1000 to 1024 : 3000	C
	AddNoise	A white Gaussian noise is contaminated the watermarked signal to simulate ambient distortion		35 dB level to 5 dB	D
	AddSinus	It adds a sinus signal to the watermarked signal		120 : 3000 to 150 : 3500	E
Conversion	Resampling	The sampling rate of the watermarked signal is converted to and then is reconverted to		4 KHz : 16 KHz to 8 KHz : 16 KHz	F
	Requantization	The sample of the watermarked signal is quantized to and then is requantized to		8 bits and 16 bits	G
	Invert	It inverts all samples in the watermarked signal, like a 180 degree phase shift	NO PARAMETER REQUIRED	None	H
Ambience	Echo	An echo with a delay and decay is added to the watermarked signal		20 ms and 10% to 100 ms and 50%	I
Sample permutations	Cut samples	samples are removed from the watermarked signal from every period		1 and 1000 to 7 and 1000	J
	Copy samples	Some of the samples of the watermarked signal are copied between the samples values		1000 : 100 : 30 to 1000 : 200 : 60	K
	LSB Zero	Set all samples of the watermarked signal to zero	NO PARAMETER REQUIRED	None	L
	Smooth	The new sample value depends on the samples before and after the modifying point	NO PARAMETER REQUIRED	None	M
	Stat1	It averages the sample with its next neighbors	NO PARAMETER REQUIRED	None	N
Dynamics	Amplify	The amplitude of the watermarked signal is increased up to and is decreased down to , respectively		150% and 75% 200% and 50%	O
	Denoising	The watermarked signal is denoised by	.	−80 dB to −60 dB	P
Filters	Low Pass Filter (LPF)	The watermarked signal is filtered by an elliptic LPF with cutoff frequency of		5 KHz to 4 KHz	Q
	Band Pass Filter (BPF)	The watermarked signal is filtered by an elliptic filter with bandwidth from to to simulate a narrowband telephony channel		500 Hz & 4000 Hz to 300 Hz & 3400 Hz	R
	High Pass Filter (HPF)	The watermarked signal is filtered by an elliptic HPF with cutoff frequency of		500 Hz to 800 Hz	S
Time stretch and pitch shift	Pitch scale	The pitch of the watermarked signal is nonlinearly scaled without changing the time		1.05 to 1 : 10	T
	Time stretch	The time of the watermarked signal is nonlinearly stretched		1.05 to 1.10	U
Compression	CELP coding	The watermarked signal is codded with rate of by CELP codecs and then is decoded to original one		16 Kbps to 9.6 kbps	V
	MP3 compression	The watermarked signal is compressed by MP3 with different rate		128 to 32	W
	G.711	The watermarked signal is codded by standard 64 kbps, A/μ-law PCM	NO PARAMETER REQUIRED	None	X