The value for f_p , f_s , and f_T are estimated from the graph,
f_s = 3417 Hz (looking from the graph)
f_p = 243  Hz (estimated by listening to the quality of sounds filtered)
f_T= 8000 Hz (return value from wavread function)
The normalized cutoff frequency is      w_n = fp + fs / f_T = 243 + 3417 / 8000 = 0.4575
The N samples = 48
The results of the filtered process is a filtered signal sound which is noise free and almost sounds the same as original signal sound. Except it is lower in volume compare to the original one. The resultants graph produced and required is at below.

MSE and SNR (after filtering) calculated using matlab code , please refer to this link MSE&SNR to see the matlab code
MSE for noisy signal after filtering = 0.0485
SNR for noisy signal after filtering= -10.7118
MSE for noisy signal before filtering = 0.0573
SNR for noisy signal before filtering = -12.3869
SNR before filtering is smaller than the SNR after filtering , which means the filtered signal has better quality and less noise.

Graph for original signal & Graph for noisy signal ( please refer to this LINK to see the graphs )

Results from filtering process.

DSP chip coefficient

Graph for filtered signal